From Data Analyst to Director in 3 Years: How He Did It - (Richad Nieves-Becker) - KNN Ep.90
Updated: Mar 24
Today I had the pleasure of interviewing Richad Nieves-Becker. Richad is a self-taught data scientist with an eclectic background. He currently leads the data science function at Revantage, a real estate shared service organization in the Blackstone family, and enjoys regularly sharing his perspective on the field on LinkedIn. Richad got a BA in Neuroscience and Anthropology and was on the PhD path until he realized a life of basic research was not for him. He pivoted and earned a Masters in Commerce (concentration in Marketing and Management) from the University of Virginia where he met me. After a brief stint in inside sales, Richad committed to data science and built his career in the real estate and finance spaces. He started at CoreLogic focusing on text mining, then moved to Greystone leading all things data in an innovation lab, before joining Revantage. Richad credits his career progress to focusing on impact and deeply understanding the business case. In this interview we gain some insight into the specifics of data in the real estate domain and we learn about how Richad was able to go from analyst to director level in less than 4 years.
[00:00:00] Richad: So when I write bullets well, first there's the description. Then I write bullets with every bullet I write down the requirement and the requirement is usually like the general principle, the broader category, and every single requirement has a preferred every single one. And the preferred is closer to our use case, for example, required, done something in cloud.
So let's say preferred Azure, because we use Azure. Right? So it's sort of making the case that like, if you don't have Azure, don't apply. But you do want some cloud where you can like converse with cloud. Talk about it, have used at least like a couple tools in something like AWS, even, right?
[00:00:49] Ken: Today, I had the pleasure of interviewing Richad Nieves-Becker. Richad is a self-taught data scientist with an eclectic background. He currently leads the data science function at Revantage. A real estate shared service organization in the Blackstone family, and he enjoys regularly sharing his perspective on the field of data science on LinkedIn. Richad got his BA in neuroscience and anthropology, and it was on the PhD path until he realized that a life of basic research was not for him.
He pivoted and earned a master's in commerce, a concentration in marketing and management from the University of Virginia, where he happened to meet me. After a brief stint in inside sales Richad, committed to data science and built his career in the real estate and finance spaces. He started at CoreLogic focusing on text mining, then moved to Greystone, leading all things, data, and an innovation lab, before joining Revantage. Richad credits his career progress to focusing on impact and deeply understanding the business case. And this interview, we gained some insight into the specifics of data in the real estate. And we learn about how Richad was able to go from analyst to a director level in less than four years.
Richad, thank you so much for coming into the Ken's Nearest Neighbors Podcast. Obviously we've known each other for quite a long time. We went to grad school together in a field, very different from data science, but we somehow both found ourselves in this domain. And so I'm excited to talk to you about your progression.
I think listeners know my progression into the field from our same program, but I really want to hear your story a lot about, you know, working in data science and real estate, as well as managing a team. I know you've had some very unique experiences there. So again, thank you so much for coming on.
[00:02:33] Richad: Yeah, sure. Happy to be here. It's pretty
[00:02:35] Ken: Excellent. Yeah. I mean, this is so cool. I think things are coming full circle and I always appreciate it when you know, we didn't talk too much in grad school. We were in different, different I guess they were like paths, right?
[00:02:50] Richad: You were in, you were in finance. I think you're in the finance track. Right?
[00:02:53] Ken: So I was in marketing, but we had like virtually all different classes. I think we might've had one overlap class, which is unfortunate, but look at, you know, we're here now, so. But the first thing I like to ask just to get all the listeners familiar with you and a little bit of your story is how did you first get interested in data? Was it a pivotal moment? Was there one thing that happened that caused this cascade or was it a slow progression into a data related discipline?
[00:03:25] Richad: Hmm, that's a good question. Well, I would start by saying that I was always the person who thought they knew what they wanted to do. So when I was in fourth grade, I saw the neurosurgeon, Ben Carson, given talk in like a class trip. And I was like, I want to do what he does. I want to do be a neurosurgeon. And I kept with neuroscience in general till the end of undergrad. I was in a lab for two and a half years. I was very much a science-y person. And that way of thinking the sort of skeptical, like how do you do experiments?
How do you take a critical mind to everything? How do you break things down? That was very much like natural to me, but also like growing up, I was just exposed to a lot of that. It took me two and a half years and undergrad to realize how much I hated the lab that I just didn't like basic research at all.
It was very isolating. It felt like there was very little room for creativity. That was one thing. And another thing was, it didn't feel very analytical, which sounds a little weird if you're in basic science, if you read science, you're like, Wow, this must be really analytical people writing this.
But actually the doing of basic science, especially in biology is very rote. It's like a robot to do it. You know, like why doesn't the robot pipette? And then they won't mess up where they're pipetting, you know? So and also it was very isolating and I wanted a more like fast paced social environment.
So then I was like, well, I think I'm going to go to this Ms. Calm thing. That's where we met. The same, same year, same time. And I had no idea what data science was. I remember they gave us an article like sexiest job of the 21st century, which is pretty funny. Okay. Thinking back on that we had like SQL for a week or two it was like this almost like a side, this random thing.
And then we had a customer analytics class and then we had a final project with Kate Spade. And I remember we were taught SPSS sort of like point and click basic ML stuff. I don't even know if he used the term ML or machine learning. He just was like, this is how you do statistics. And like, here's some PCA like PCA clustering.
This will group the data together. And I remember thinking, Wow, it's so fascinating how you can take the same piece of data or pieces of data. Cut them in 50 different ways or depending on the algorithm you choose or what you, what you choose to do with it. I didn't even use the term algorithm back then because they didn't teach us that you will get different results.
And I was like, so there's a million ways to look at the same thing. It felt like magic. It felt a little bit like magic to me. This was all in reflection later. So in the moment I liked it I think when we made a presentation to Kate Spade was interesting. I think our group, our professor really liked the stat stuff, which I did all the stats of basically I think that the only a hundred percent in the program, but we were not picked by the CMO of Kate Spade to go onto like this to like present at this like big thing.
And that was a lesson for me, actually, my first lesson in business versus academic. Like, what do business people care about and what did the academics care about. At that time, I still didn't want to do data science or become a data scientist, start anything. I joined. I became an inside sales rep for a sort of learning and development psychology company.
I was there for four months. I learned a lot about cold pitches and emails and but I had an awful boss and made you feel like a number, I suppose. And so I quit on Christmas and that was the end of 2015. And I was like, Ah, what am I going to do with my life? What am I going to do? I was like, maybe I should start a business.
Like I was like, Oh, I'm interested in neuroscience and sleep. Maybe it will teach me a lot of sleep better or something. And during that time I reflected back on all the things I had done in school. And I thought, well, maybe instead of just focusing on the purpose of the company, I should focus on enjoying the day-to-day activities.
What was the most natural to me? What was the most natural in my experiences in school? So the stat stuff then started to stand out. It was like really obvious, like when we did the SQL stuff in school, I remember like, it was very, it felt very obvious to me. I was like, Oh yeah, of course. That makes perfect sense.
And then I remember sitting outside, like the place where we're taking the test on the computers and like the MacIntyre lab. And people would like the sort of the circle appeared around me, you know? People were asking me questions and I became like an impromptu study guide leader before the test.
And I remember that, and I remember like how natural it was to use SPSS and thinking how fun it was. And I was like, I should be a data scientist. That's when I decided, so that was probably like early 2016. Then I couldn't, I didn't have a degree in the field. I didn't have a stats background. I couldn't get an interview to save my life.
Resume submissions. I'm pretty sure I got zero interviews from resume drops, but I said, wait, I have cold sales skills. So let me research these companies and let me like backwards engineer, their email addresses, which is a skill we learned in that job too. Like, how do you guess and check like the email format of a company it's like an art and art and a science and then write a pitch.
And I had like a couple templates that I made. And then I would experiment with the templates and like do good research and okay. What's the most relevant thing I did or how can I use the keywords from their job description? And I do that and I got a one out of three response rates. And I got a bunch of interviews from that.
That was cool. I failed the interviews. I remember getting a basic verbal SQL test, which I did not know was a thing. And I was like, it was a very basic, like, how would you join these two tables? And I'm like, what are you asking? And they rejected me like, Oh, I'm not sure you can upscale fast enough. And it was like that and like that, right.
Then something lucky happened. So I was learning, I probably would have eventually stumbled on a job or learned enough or been like, Hmm, maybe the interviews, the weak part. I my mother met someone on a vacation. She went on who happened to be the chief data officer of a company in real estate.
And I didn't have a particular interest in real estate, but I was like, Hey, ... Right. I was like you see this person you met, she's like the head of sort of the field I'm trying to get into in some sense. And she's like, what? Really? And so we ended up meeting for lunch once just to meet me.
And at the end of it, she was like, yep, you have a job, some kind of job. I'm like, what really? Okay. It took a while. She introduced me to three executives under her. They all interviewed me trying to figure out what I could do for them. And they, none of them had anything for me. So it sort of sat for another like month or two, I think.
And then one of them. Got back to me and he's like, Hey, I created this position. So sort of, you could imagine on the backend there's strategy discussions and like, Oh, we need to move more into this like ML area we need, like, you know, entry-level analyst types. And so they're like, I made this position, had you in mind, it's called data research analyst.
And I was like, cool. And I interviewed with four or five people, and then I got the job. So then I moved to California Irvine, Irvine, California. So, so Cal very nice weather kind of bittersweet because my now wife, but then girlfriend, you know, we were in distance for a while. We're on the East Coast.
She was not gonna move to California. It was a good chance for her to see if she liked it, but that wasn't gonna happen right away. So now I was going to be cross-country relationship, but that's how I got into data. And that's how I got into real estate data to, I ended up learning to really enjoy real estate data science because of the scale of the decisions that are made with it and unique and interesting data problems about data quality, that's getting crappy data quality that costs still costs millions of dollars. Yeah.
[00:11:49] Ken: Well, you know, I love that story because although you didn't necessarily land a job through the process you were using, you still had a process for approaching, interviewing and looking at the bottlenecks in a very systematic way.
Right? You said, Okay, I'm not having success in this more traditional way. What are some other approaches that I can take? Can I test this? Can I evaluate my progress along the way and see how I'm doing? You just described, Oh, I had a one in three success rate with this other method. That means you are tracking your progress.
You were doing data science on your process. And to me that is something. Like even in an interview process, if someone told me that that would be like, Oh, that's really good that this person thinks in this way. And you can think like in your daily life, different from a data scientist and in your, in your work life in a different way.
But I generally find that the successful data scientists that I interviewed her, that I talked to their mind works in a certain way, or they've trained their mind to work in a way to solve problems with data. What, I mean, look at me and my personal life. I track my sleep stuff. I track everything I eat.
I do like it. It's just a way that I think, and that I go about solving problems. And I think that a lot of people there's this discontinuities between the job search and the actual work that they do, when if you zoom out, it's the same thing, right? Like you can use data to figure out your job, search, to land a job, to improve your skills, to do any of these things. And if you're a data scientist. Kind of a little shame on you if you're not doing that. Right.
[00:13:33] Richad: It's really interesting. I haven't had enough conversations with current data science aspirants, where I propose something like Hey, why don't you do this instead of that? and then hear their objections or, Oh, it didn't work or whatever to know whether like in the end, what is their motivation for getting into data?
For me, it was a way of thinking. And it was, I guess, like a matter of time until someone would see me and be like, Oh, this person, they have like potential because of the way they think and their persistence. And maybe some of the creativity aspects too, of doing something like that. But. Yeah. I mean maybe with some others, they just do it because it has a salary and it's like, looks cool.
And they're like, Well, you need a job or something like, that's not a good motivation because the sort of thinking and also data science and practice, it's quite, it's quite challenging. It can be it's the only, it's the only major function in a company with no guaranteed ROI. Right. You, everything we do is research.
That means you don't know if it'll work, you know, that's and that's something that a lot of companies don't understand or that's something you have to communicate and it's something. I think about when I started, I just thought all these, these the stuff's cool. Right. But if you're thinking about like, I'm a company, I have funds, I want to like distribute the funds to do useful things and compete or whatever.
Then they're thinking like, what's my return going to be on this. And that's, since I guess the hype makes them want to spend, but to actually like do it in practice, it takes like an intense sort of systematic research, focus and creativity within that framework.
[00:15:16] Ken: Yeah. Well, there's sort of like a meta research, right? Where you have to build models and simulations and projects, how much you can expect to make from any given project. Right. And it's about expected value, not about actualized return, which can create some really interesting accounting implications because you're talking about cost savings over the course of next year from data science, operations, and it's not actualized cost savings.
Right. It's like a hypothetical cost savings and you do that, right. It's a very interesting, I was working at a large company and that's like, one of the things that the data science teams were doing is we were like estimating savings from the analysis. Right? Yep. And it had real dollars and cents implications on taxes, but it had no real dollars and cents savings in the real world, or at least that had been actualized yet.
And I don't know, like to me that was just like an absolute, like brain noodle type thing. Yeah.
[00:16:12] Richad: Yeah. That's super interesting. Like what do you tell the accounting when they have to like, yeah, I don't know if it's capitalization versus I don't know what it is, but that makes sense.
[00:16:22] Ken: And so walk me through you land this first job.
How did you get to where you are now? You had a pretty aggressive assent into leadership and running your own team and doing those types of things. You know, from talking to you, obviously. I think it is, you know, one of the things that really helped that is this multidisciplinary background that you had.
I mean, you talked about sales, you talked about kind of the problem solving. I mean, at least until you landed that first data role, I would say that your, your like pure data abilities were probably secondary to the other skills that you brought in. Probably. How did you upskill in the like, data stuff.
And then how did you balance those and make yourself as marketable as possible for, for leadership and growth?
[00:17:06] Richad: Yeah, that's a good one. That's a good one. This is basically like the next step of my career. So I joined, I joined the job. I am like, Cool. This is my, this is my desk and my manager was there and she was like, Oh, this is like, these are the people that are around you.
and here's like some of the data that we work with and we sell data, basically they sell real estate data's core logic. It's one of the premier, I guess real estate data companies. They basically specialize in public data in the county offices all around the United States. So every time you like a property has a transaction or something like you buy, sell a home or something, it gets recorded.
And then the counties also assess the value of your property for taxes. So those are thousands of different data sources in pieces of paper. And these forums, you know, manually gathered in order to get this data into like a database, a structured format that is then sold to. So it was other things too, but that's like, that's like the meat of it.
And that was, I was on the MLS focused on MLS area, multiple listing services. So when you try to sell a home and you put it online and you write a description like, Oh, beautiful, two bedroom, two bath, blah, blah, blah texts. And then you put in like the beds and bats and you can search for homes that fit these price ranges of characteristics.
That's like that that's the data that I was working with. So basically they showed me this data. They're like, we sell it. And then and that was it. And I was like, Hmm, what should I do? I was also had a career conversation early on with my manager's manager. The one who brought me in initially and he was like, you could go one of three directions.
You could be like a business analyst. The requirements and all that, you can go full like data science ML, or you could go in the middle, it'd be like a sort of data analyst type. And I was like, well, I know which one I want to do. So basically I had, they had really no idea what to do with me. They sort of, I guess, figured that I would figure it out.
So this is what I did. This is how I came up with my own first project. I saw the data and I was like, well, look at these text fields. There's probably lots of interesting information there. I also saw that a lot of their major structured data fields were no. And I figured that I could use the text to impute values into the structure data.
We could extract information fairly basic, like texts use case. I knew very little. About it at the time I was halfway through at Xs MIT course, the analytics edge with our so I started with R I chose R because of it had a better visualizations. And they said it was better for someone with the more of the stats side rather than the computer science side, which was definitely me.
So I said, okay, I'll lean into visualizations because a PhD might know a lot more math. I was the only non PhD person there who was doing ML of any sort. I was like, Okay, I will compete by focusing on vis and I will focus on the business use cases and asking like good questions. That's pretty much what I told myself when I joined.
So in order to make that first assignment, I sort of put two and two together. I was like, here's this text? Let me use R to see, and then pick some fields. Like I read some of the texts manual. And I said, Oh, it looks like they described things like pools and fireplaces, and sometimes they say the year built and these are fields we care about.
So what I did was I took the text and instead of doing ML, I did not do ML. I did some stemming of the words. I made a word cloud to see the common 50 most common words. And then I saw that some of the 50, most common words were related to some of the data fields that we want to do. Impute. Like they're literally present like fireplace or pool or something like that.
I figured people would do would write fireplace and pool because that's a valuable thing when selling and buying a home. Right. So I saw that, I showed it to people. I'm like, look, these words exist. Let me do ML to like, predict the presence of these fields. And they're like, whoa, what? That's cool. So cool.
Like they had never seen anything like that. And they're like, Go forth my child. So, and then, so that was my project for a year and a half more or less until I left. Why did last year and a half? Well, that's, that's a different story, but a lot of platform changes. So I'm like, I guess I'll rewrite this in Scala spark.
And then I guess I'll try this algorithm and, you know, it's sort of like the epic made me aware of the context around data science that can make or break the actual delivery of a project. But for me, as a first time an analyst role, it was a lot of fun to like, just learn more ML, try to improve the accuracy, what I was doing do feature engineering and basically educate myself on the job.
But I guess the key, the key difference is that I knew I had to figure out my own work and I also figured that if I made the case for my own work I would, I could pick something that I would find more enjoyable and interesting. So I picked texts and unstructured. We're semi-structured I guess. So that was, that was CoreLogic. And then...
[00:22:24] Ken: Before we continue, I want to highlight something that you said there is that you like thought about within your own work, where you would create your own competitive advantage. Right? I think a lot of the times as data scientists, we think that we have to maximize in every category.
Oh, I have to be the best programmer or I have to be the best at XYZ. I have to be the best at this type of analysis. There's tremendous value in specializing, not just in like the technical dumb. But across the skill set that you have. So you really did flex the visualization side of the things that other people in your, in your category or in your company might not have invested as much time in or be willing to work on or whatever it is.
And I think that that's a really powerful tool for advancement, right? You don't not, everyone has to like purely advance as a individual contributor data scientists. Like there can really only be one tech lead. There can really be only one lead data scientist on each team. There's plenty of other roles within the domain that you might really like and might really be suited for.
If you're thinking about your skillset and what you bring to the table to begin with. I mean, something I really like is that you, you didn't say, Hey, I'm going to come in and become the best programmer in Python or whatever, the most popular languages you said, Hey, let's look at my history. Let's see what the next logical step would be.
So our made sense for you because you're like, Oh, I come from like a little bit more of like the statistics background than the CS background. I wanna make sure that you know, I did research into the visualization tools. I actually don't think are still has the best visualization libraries.
That GG plot is really good, but I can do GG plot on Python now. So but I mean, I digress. I think the idea though, is that you, you married your history incredibly well with the future that you wanted to create, or like the projects you wanted to create or the skill set that you wanted to build to advance in the field.
And I don't think we do enough introspection about ourselves compared to saying, Oh, like, this is in super high demand in the market. Like if we're looking at, at, you know, job perspective, right. I'd say 80% of roles prefer Python over R right. But if your background. Is a better fit for those 20% of our roles and your story is better and you have a higher percentage chance of landing those roles.
Like it might really make sense for you to pursue our more heavily than Python, even though Python at a global level is the more popular language. So, you know, maybe I'm extrapolating, but I really wanted to highlight about that kind of angle of your story, because I think it's so cool.
[00:25:04] Richad: Yeah, if you want to answer to get to the point, if you want to know how I was able to reach the director level in data science, go from analyst to director and under three years, the one word is specialization nation specialists.
Maybe that plus mindset, but it wasn't definitely, wasn't a track record, a stoning track record of, of like success and deliveries. It wasn't that, unfortunately. So I guess continuing the story, I suppose I w then girlfriend at the end, I decided that we liked New York, the vibe of New York city, better than the vibe of LA or, and so Cal.
And so I started looking for, to move back to the east coast thing is, it was still impossible to get interviews at the time. I don't, I'm thinking back, I don't know why it was difficult, but I was like, well, I guess maybe I could do cold emails or something like that. Also. I decided that time to try to lean more into like the side project thing, like do a side project I ended up picking something that was like too big to chew to do, like after work every day and end up just feeling like frustrate.
And then on a whim I decided to look for, and I wasn't sure, like, Oh, real estate, do I want to stay in real estate? Like I had a real passion and still do for music and music, data science was just a fascinating idea to me. I w I thought at the time, like it'd be really interesting to come up with a better music recommender system because I didn't really, like, Spotify has recommendations for me especially in the hard rock, heavy metal area.
And I was like, well, I was thinking back to like musical tastes in my family. And I was like, what factors is it? And I feel like it's not just like the typical recommended stuff. I was like, Okay, maybe I'll do something related to this. But I was like, you know what, maybe I'll just apply to a real estate data science job.
So I looked up on LinkedIn. I remember there were two of them, and one of them said senior data science. Real estate stuff. I submitted a resume and then it got picked. I couldn't believe it. Meanwhile, I was applying to like junior level or like basic entry-level data science stuff in other fields and getting nothing from it.
But here is like senior data scientists, real estate. So I interviewed with them. They flew me to New York. I talked with the guy the day after Christmas in the city, it was very cool. The interview was supposed to go one hour. It ended up going to, and by the end of it, we were talking about our shared love of Alice in chains and singing and playing guitar and things.
And then at the end of. Yeah. He's he's like, so, blah blah. He's asking like logistical type questions as if I'd gotten it. I was like, huh. Interesting. You didn't ask me a technical question yet. And he was like, Oh, huh. So he asked me one thing, I answered it. It was good. I got that job. Ironically, so that was December a month before I got an interview at capital one.
And also went to New York for that. And I failed that interview and it was for 35,000 less dollars, naturally 45. It was for 45,000 less dollars in the position I got. So sometimes, sometimes it pays to, to pick a niche that is underserved and to like really go down that channel, right. In this case, real estate, real estate data scientists.
Even today, very rare. Like no one really has this experience. Real estate also has, there's a great premium on understanding your data really well. It's not like it's not like say images or like audio or something where it's like, well, you can get enough data. Like, you know, the picture quality, like you see it right with the real estate.
It's like, how do you know that this data is good? Well, it's you don't, you usually you have to go to the property. You have to like trust the source. This is a big deal in the industry. And that leads to like quality issues. And in real estate data also, it's almost never generated as a primary source on like other many other fields in data science.
Like you don't generate your own data, like say you do with the ad tech or something. It's just, it's different. Right. So I guess in my case, like where my technical skills like amazing or incredible, no was I like a gifted programmer and not really. Did I care about the business outcome? Yes. I could speak that language.
Like I tried to focus on measurable results, like listing measurable results and things, and the scale of the data I worked within my resume. But ultimately it was the real estate plus some tech stuff that did it. If it was R, Python, it probably wouldn't have mattered because it was like real estate.
That's what it was. So I got, I got the job moved to New York, eventually got married. So that's where we're here. Here we are. Yup. And then as far as, and then it was one more job leaps. So the executive, the chief data officer of CoreLogic she left to go here where I am now. And I remember calling her up for advice cause I was like, Oh, I'm applying to I'm going, gonna be interviewing for a director level role.
And I really want to like lead it to. Like, I like to mentor and and, the like, and I, but I don't have much experience in it. So I'm wondering like how I should frame myself and how I should talk about it. Right. And she gave me the advice and at the end of it, she's like you know, I might have something for you, like, keep your ears peeled, something like that.
And I was like, huh. Okay. And then that eventually led to where I am now. So I was basically like the founding data scientist of, of where I'm now, which is where vantage it's part of the Blackstone family. Blackstone's more real estate assets, all the real estate, more assets under management, any anybody else on earth in terms of across asset classes and geography.
So it's like all the commercial things like multifamily, like I'm in an apartment building now, not Blackstone, but you know there's office, there's retail, industrial, which is a hot commodity because of shipping, you know, you need more like nodes of the network in order to deliver goods to your home, you know, do Amazon prime, that sort of thing.
And and other things, yeah, there is hotels like there's many categories and they, they, and we figured, Oh, that's a neat opportunity because you could do, you know, all sorts of analysis that no one else on earth could do because we have access to all this data. The reality is challenging in terms of doing that, like the legal implication of doing that in the structure.
But that was sort of what appealed to me in many of the other data, people who join it's like a data playground. Right. So when I came in, I guess the final, final thing I'll end, and I think this really speaks to the. A mentor and someone who believes in you? I, so I had like a salary, senior data scientist, and then I was like, Oh, I guess I'll give like a higher salary by this amount.
And then so I gave that, I was like, Okay, this is what I think it should be. And then I get a call from them, like, Okay, you have the job, but actually your salary level that you gave us was too low, given the level you're coming in now. I didn't know what level I was coming in. I didn't even know like levels were really a thing in corporations, like large corporations.
So I didn't think to ask that. And so I was like, Oh, okay. And so we're going to bump you by this amount. And it was like, it felt like an enormous, I was like, what? That really? And I was like, giddy, it was cool. I was like, that wouldn't have happened. I asked like, why did you do that? And she was like, Oh, you just gave the wrong thing for your level.
And I was like, Oh, I'm coming in at that level. Okay. Well there, so that was like really I don't know. It sounds like, look, it is luck. Okay. But it's not entirely blind luck. It's like a luck that you have prepared for in a sense. It's like you have a strong network, you deliver well. And people like, if they believe in your potential, they will take risks on you. And so that's how I got how we got to our advantage.
[00:33:30] Ken: Well, I think that there's, that underlying message that you just described. You know, people forget that the interview process is about people. You're usually talking with people. The vast majority of interviews are not like the ones that the fan company, right?
The van companies, they're all as quantified as possible. There's four or five people that you talk with and then you're scored and then you're they evaluate if you're a good fit based on that, right? For most companies, the interview process is not like that. It's very similar to what you did. It's in this day and age where data science roles are still relatively new, there are still companies that you can be a founding data scientist on.
You're going to have to convince them of your skills or your ability to manage their problems more than you're going to have to convince them that you can code in Python or do some of these other things. And the idea that you're coming in and you're bringing subject area expertise, like in real estate, right?
That's going to be so much less foreign to them. That's going to be so much less intimidating to them than saying, Oh, we have to train this code. To understand how to put, how to, how to work in this domain. I mean, for me in sports, right? It's like if a client, a golf client, right. And they had to come in and explain what golf is to the data.
People we'd have an issue. Right. But the fact that I've played the game, the fact that I played competitively and had some experience in that, I've talked to these people before, that gives me so much of a leg up because I understand the subcultures. I understand like some of the idiosyncrasies, I mean, for example, in Chicago, right?
With the real estate, their newer buildings all are mandated to have white roots. Right. A very simple thing. Like a data scientist knows from an image recognition stance. That is a unbelievably valuable tool because you can use normal satellite data and understand what areas of the city are developing relatively quickly.
What it's looked like over time. I mean, there's a lot of information you get from that, but like someone who doesn't have any inkling about the domain there, they would be completely at a loss for how to evaluate some of these things like your feature space from understanding the domain expands infinitely and and, people and companies they can grasp that immediately see the value of people that have that domain knowledge.
So I think your story is so cool. It also shows the mentorship aspect right. Of finding, I wouldn't necessarily call it mentors, but people, Oh my God, what advocates for you.
[00:36:02] Richad: Advocates, yeah.
[00:36:03] Ken: Which I think is unbelievably cool. You know, it seemed like the lucky part of this, right. Was you meeting that person right?
The aunt, the not lucky part was you cultivating that relationship and maintaining it and having a good enough continued relationship with this person, for them to offer you a role in the future. How do you create those relationships? This is something I have really struggled with in my life is like, I really struggled to find people that I look up to that I want to maintain relationships with.
And maybe that's because I've been working largely for myself for the last five years. But do you have any advice on that maybe for me or for anyone who's starting out in their career?
[00:36:49] Richad: So, so I'm going to repeat that, finding, finding mentors, and then knowing how to cultivate and keep those relationships open and precisely.
Yeah. Okay. Hmm. Yeah. So if I thought of it like that, I would never do it. I will say that's the first thing. I'll say. The idea of if the idea of networking and the idea of like keeping your network warm or something like that sounds so calculate. But not in the cool data science way, like in the opposite way that I would never do it and did never do it.
I basically would talk to people. I genuinely wanted to know how they're doing and also like give them updates and and, things. This is a tough one because I don't know if I've quite created a model for how to build mentors. I would say that you have to figure out what naturally motivates you and then tie regular communication with people to that thing.
That's probably my biggest tip. So for example, if you really like to learn things and you're really curious person, then instead of thinking of it, as I'm going to find a mentor, a network, and then I'm going to contact them every six months, like a schedule or something you might think, well, what are some really, what are some things I really want to live?
And are really interesting to me. And then how can I learn that as fast as possible as it turns out, you can learn a lot faster. If you have basic knowledge of a thing, you can learn a lot faster by talking to a human being than you can by like reading something. And you can learn a lot faster by reading something, then you can by like trying and failing.
Although you end up learning a lot more from trying and failing. So there's a room for all of those, like trying and failing on your own reading. And then like talking to human beings, data scientists tend to lean the least on talking to human beings. And it's probably thus a great area of, of differentiation if you're like that.
So if you can figure out how to make yourself do that, then you'll have like an enormous leg up on, on people. So I'd say like what, when you look at your week and you're like, what are the things that were the easiest and most natural for me to do? And then why did I do those things? And then how can I tie talking to people to that?
So for example, like I did this thing, LinkedIn hard mode. It's a challenge. For data people, basically, it was sort of a group cohort use the hashtag and you post everyday on LinkedIn for 30 days. There's certain rules like,
[00:39:16] Ken: And that started by Al Bellamy. Right. I want to make sure I give Alan shout out. He shouts me out way too much.
[00:39:23] Richad: Nice. Yeah. So this challenge, there's certain rules, like all original content and no poles and there's like other things, but the most important thing is post every day original stuff. And so the cohort goes through that and then they learn a lot about like themselves.
When I, when I did it, I had the goal of, I didn't really have the goal of like, Oh, I'm going to do this every day, come hell or high water. My goal when doing it, the way I thought of it was what's a, I'm going to discover a way of writing that makes me want to write that. And I want to discover a way of doing this that pulls me to want to post and write every day or write posts and batch and schedule, which I often did.
Which is good too. So I w by doing that, I was able to like experiment and figure out like, Oh, how do I use less self-discipline and get more excited to do it naturally. And as it turns out, you post on LinkedIn every day, you build this like big community, and then you start conversations. And if you're genuinely curious about the conversations or to learn more about those people, you'll message them doing the challenge.
I ended up messaging so far, what, over 150, ah, probably at least 120 and the 98 people I've had. I've had conversations, text conversations with 98 people that I never met before which is like just pretty good for a month. That's like a lot of messaging and after work, too, you know, so like Netflix. If you think of it as networking, or like, I'm going to get something transactional that that's not fun.
I wouldn't say that like, Oh, that makes you calculating. I mean, maybe some people would be motivated by the idea of doing that, but for me, what makes learning data science more fun is to create a bigger, a bigger community of people. And to make it more like serendipitous, like learn stuff. I didn't even know.
I didn't know, like be encounter something I'd never seen before and like, wow, that's cool. Like that's keeps it fun. It keeps it fresh. If you tied network building activities to that, you'll, you'll be really motivated as far as like, and then as far as finding a proper mentor, so instead of the community, talk about mentor.
I would say the more yourself you are, the less you're trying to put on airs and like get I don't know, you're like targeting a specific person and you really want them, and you're going to like change the way you come across to do that. That's like the wrong way to do it. If you're you're, if you have these conversations and then you are naturally, you are your natural self and you show your natural enthusiasm for the things that you actually care about, the right person will see that you are uninhibited and then they will really want to mentor you.
That's, I've generally, that's how it worked with, with mine. I advocate but it would probably what it would do is eliminate a lot of wasted time. If you don't do that, you might end up wasting a lot of time with someone who's not right for you.
[00:42:19] Ken: Yeah. I, you know, I think my views on mentorship, I've always been like, jaded about the transactional nature. I mean, maybe that's the position that I'm in, right? Is that a lot of people say, Ken, will you mentor me? Will you do this and that? And I'm like, look like for the most part that the time exchange for mentoring someone where I'm at now, I can't justify it. And I do it in the best way that I can through scale, which has videos and talking to things and webinars and whatever it is.
And I find it hard to think that, Hey, if you know this person who I would look up to, you know, I generally would want a mentor that is significantly more successful than I am. How could I impose on their time? I don't think that that would be fair of me. And I, you know, it's interesting in terms of cultivating friendships and those types of things.
Like I have quite a lot of friends that I look up to for advice and thoughts. But I w you know, there's just this weird, like mental bridge for mentorship or advocacy that I haven't quite understood how to traverse yet. And I might not ever, I mean, part of my mindset is that, like, my goals are almost always to grow past everyone that I meet in a positive way.
And like eventually have them, like working with me on projects and looking at me as equals. And maybe that's like one of my competitive advantages too, is that I use that as some weird form of motivation as well. But it it's an interesting way to approach it. You know, maybe I should be looking at these friendships as like like almost a a loose form of mentorship where it's like a collaborative type thing.
And I've been looking for something in like a pure mentor that probably doesn't exist for where I'm at in my career, especially with a lot of the stuff that I'm doing. I think. A lot of people haven't, you know, aren't in the space, you're limited by the size of the industry or the size of the content space or whatever it might be.
But really interesting perspective and giving me some pause and some, some additional thoughts here as well.
[00:44:34] Richad: Yeah. I'd summarize it as make meeting and talking to people fun for them too. That's the biggest, yeah. Mutually, mutually. Yeah, you could think later about how to make a transactionally mutually beneficial, but you just make it fun.
Then people will want to take time out of the day to talk to you because if you leave them better than you found them, and they're like, Oh, that's interesting. Like I talked to people, you know, mentors or people ahead of me in career, whatever. And I'll just be like, Oh, I read this article. I like saw this thing and I'm doing that.
And then like, just like the stuff I'm doing, it's like exciting. If you just, you know, Oh, this data thing and, Oh, can you believe that this politician did that? And I don't know. It's like almost like a friend almost.
[00:45:19] Ken: Yeah. Well, I mean, that's one of the, my favorite reasons for reading quite a bit.
Yeah. I'm probably on pace. I read about 50, 55, 60 books a year. And that's something I can communicate with almost anyone on where even if they're significantly more successful in their career, they're doing crazy things that probably means they have less time to read and absorb new information. Right.
And so being able to summarize a book or tell a story about, about something that might be specifically relevant to them, I think is a really cool thing. And it's like, well, yeah, I mean, you're busy, but Hey, if you make time to read this book, it's perfectly highlights this unique situation that you're in.
Speaking of books, there's one I do want to recommend related to something that happened to you, I guess, in the last couple of years. And so you, you landed this new role where you were leading a team. And then from what I understand, a lot of that kind of. Like, you know, like it turns to sand and falls through your hands.
Can you tell me a story about, about what happened and about sort of leadership and what to do if a team you know, it was underperforming or it doesn't go to plan or, you know, politics get in the way of these teams.
[00:46:38] Richad: Yeah. Yeah. So essentially what happened was I joined my place. I grew the team to four, including me. I, so having now known enough about data science, I realized that I wanted some people. The ML experimentation, and then there's the stuff like stakeholder seat, right. There's a related they're related. And, but then there's like all of the infrastructure stuff in the backend and like the sort of back end experimentation, that's very scalable in terms like, what tool should we use?
And like what long-term investment should we make in order to make data science and enduring function. So I wanted more than just me and I wanted more than just two. I wanted four and I got four. And then things were, seemed to be going pretty well for like a year and half. And then in at the end of 2020 basically I learned that my team is going to be cut in half.
This was like a decision to this day. I don't know who made that decision. No one has told me no one, no one in my company. It was like the parent, the parent. Right. And that was pretty hard. I learned Monday and I couldn't tell him till Thursday, which was freaking awful. I feel like an asshole to something like that, but this is the discretion required of you sometimes in different environments.
So what happened then? So one thing I did, I had a conversation and I said okay, it looks like we're hiring more data engineers. In next year we have open requisitions for them. What if we give the people that we are letting go and an offer to stay on as data engineers. And it was very easy to get everyone to agree to that, actually.
So we were able to be like, Hey, you can take a package or you can stay on as data engineers. I predicted accurately predicted who would leave and who would stay. So they, they did then the one who stayed it turns out they were wanting to move into a pretty thorny project that he just didn't want to do.
He wanted to do. Something else that was engineering related and a bit more innovative, I'd say not like a, like those quagmire projects that are really important and have been like a big money and time sink and are sort of complicated things done Tang. He didn't really want to do that. Okay. And so he resigned.
Right. So then it was just the two of us. And so what do you do in that situation? Well, I can tell you the mistakes I made and I can tell you what I probably do now. The mistake I made and it's probably how my brain works. You might notice that if you think about my story about, of getting into neuroscience and how long it took for me to realize, I didn't actually enjoy the work that emotional awareness, emotional awareness, or like, how would I feel doing this?
Can sometimes be disconnected from what my brain will naturally do, which is come up with a great strategy, a great sounding strategy. So I basically came up to like, Why don't we do this kind of work? Instead, the thing is the thing I came up with everyone was like, that makes a lot of sense, given the reasons that the team was like go and what the they're probably looking for and stuff.
Everyone said that it makes sense. I didn't want to I'd realize, I didn't realize till later I didn't want to do it. I don't want to actually do that thing. And so I didn't see, I wanted to, I wanted to be more customer facing. I wanted to serve PCs portfolio companies. I didn't want to like just do internal sort of tool building, you know, I didn't, I didn't want to be that person.
I wanted to like. In the real estate data itself in the financial data itself, the customers, as, as a shared service, our customers are the real estate portfolio companies of Blackstone. Right? So not like business to consumer, it's more like B2B, but all within the private equity blanket. Right. So how did we eventually turned around?
Well, we worked hard to find one project on which we could put a really solid ROI case and time-saving case. So I was like, Okay, I need to find a project that will work. And then I also actually want to do, and I thought, well, here's a project, a forecasting related project that could save people time, save their time because they were doing a manual forecasting process, a lot of art and science, a lot of art to it as well.
And I'm like, what if we use like ML to help automate that and scale it so that they can spend more time focusing on accepting. So it's not taking humans out of loop. It's more like scaling their ability to do things. So they don't have to pay attention to the business as usual stuff, and they can focus on exceptions and their forecasting work.
Right. And so we propose that and we're like, Okay, we learned from them, like, these are the number of hours per month that they do this over this many days. And so we basically made the case that this project would help save them late nights that they are pushing right now. And it would, and it would reduce the hours by X and then as, and it would take us this much time and you get ROI by this time.
So we put that essentially in in this Excel and we submitted it and then it got approved and it got approved almost like without much fuss or fanfare, because it was much lower budget than other things that were being requested in the broad scheme. So it was like, well, yeah, of course, yeah, go do that fine.
There's only two people anyway. And that's how we got like, official. Approval to do things, to do a thing. And so that's what we're working on now. So the way to turn it around is to really is I guess, to sum it up is that and I don't know if, I don't know if they teach this in MBAs or if any, if all like the execs or VPs or whatever watching this would agree, but I think need to come up with a strategy that not only will work in a, in a broad sense, but will actually be, you know, part of the strategy is yourself and the motivation of the people who are doing the work.
So you can't just come up with a strategy that no one would be motivated to exit. Because it's like antithetical to the reason they joined in the first place. Right. Something I didn't think about at the time, cause I was just feeling down about it, you know? That I'd say that's like a, that's a big lesson.
I mean, people, I think strategy comes from military, right? That's where like strong strategy ideas have been originally developed and you don't really need to worry about whether people want to do the thing you just, you know, so people notice only think about like internal motivation as like a key element of strategy.
But I think in in any setting involving people with free choice I think, I think that's a really important part and I think it often gets missed. So I would say like, you need to find that overlap of a strategy that you want to do and a strategy that would make sense in the environment. Make sure you make the real strong business case with numbers.
For the thing you want to do which I couldn't have done if they had pulled the plug like six months earlier, but we knew enough from the work we had done before to make this case. And then you can iterate from there, from there it's successful delivery and then like showing everyone your successful delivery and then you use that to argue for more delivery and then more people. And that's how you scale, essentially.
[00:54:09] Ken: I love that. And so I think that's like the technical thing that you do to turn things around, how do you turn things around from a cultural thing? Like, I would imagine that your, like your morale was low. The other person you're working with are really low. It's like, Oh my goodness.
You know, w we just had to let go or change the role of two of these people, you know, is that something that it's like, Oh, you know, we need to create momentum. We need to find something that is, that is actionable and build up and eventually like create this system from the ground up. Or is there some other nuance there around like the. I guess like the human element of turning things
[00:54:48] Richad: around. It's a good question. Well, the first thing is that you're always really honest. I, yeah, you don't sugar coat. You don't also catastrophize either. It's like be balanced, but communicate everything, hide nothing including how you're feeling.
Probably people are seeing this with disagree or maybe they think, Oh, my environment's different. We can never do that. Well, I would say I would never work in an environment like that in the first place. And that's not the environment created on our team. Like no matter what their corporate level was, I spoke to everyone basically in the same way and sort of assumed that the same, like autonomy and like put the same weight on what they wanted to do.
I'd say that the main, the place you start in a turnaround or like a down situation when you have that is that you get really clear on each individual's goals. It's like, it's not like you do one thing and then the next. You were working on like a strategy or a way to bring yourself out of it.
At the same time, you are talking to people, having Frank conversations, where they can, where they feel vulnerable and able to tell you how they feel down to like, I am interviewing, you know, or something like that. Right. And then you you're like, Okay, what are you really looking for? What do you really want in a career?
Like, and then you think what's the best way that I could align what they really want to like the strategies that I'm coming up with, you know, or that we're coming up with talking it through. So it's really about finding the alignment as you might imagine. This is a creative exercise, involves a lot of creativity.
Another thing that so that's like a general answer. I think that applies in any situation. Something that I did. So the person, the last person left under me eventually left and in March and I'm like, no one blamed him. Like, no one was like, why did he leave? Like, you know, so after that I said, Hmm, how can I make this more interesting.
So I decided to try to, I thought I would like to see more people of color and data in data science. It's something I've thought for a while. I like mentoring. I like college programs. Like I like talking to students and things like that. Teaching and like the next generation it's it's great. So I was like, how can I change our talent pipeline when I back this role?
So that it's not just like, whoever happens to see our job description in Chicago, you know? No, I'm not blaming like any anyone like here for this, but like the candidates they'd send me there was like, 15 like white males and one female. I remember interviewing and I was like, Wow, I counted.
I counted. I was like, Okay. I see. And it was like, Oh, they're not out there. Well, I guess we're not going where they are because they are out there. We're just not going there. So I thought, well, what if I bring them to me? So I decided to run an, a sort of online event. And this was like my foray into LinkedIn and into the community.
This was like last year in 2021 in may I did this a five day, like promotion acceleration challenge for data scientists of color. It was basically focused. I interviewed data scientists of color and I basically, it was trying to took the common challenges that they have and what they said.
And it was like, Okay, this is how you, like, make it the business case, communicate with stakeholders. Some of the things that, you know, might be missing in order to get promoted more quickly in your own roles. And on the fourth day of the five day event, I was like, by the way, we have an open position, so feel free to apply.
And they were like, Oh cool. You know, so I sort of wanted to like lay it on like lay on the value in a sense. So to make them want to come here and do an event. So that was like, that was a huge booster. My morale to put together something like that. When I proposed this to our CEO, he was like, this is one of the best ideas I've heard in my career.
I was like, really cool. I thought that was pretty cool. Because it was so like creative and out there. And so it was really like sometimes the way out of an authority situation is not to take little steps because the little steps if you're in mud and you take little steps, you might just dig a little deeper.
Right. Sometimes you need to like lift off a lot. To like launch yourself out of there. You need to like bring more energy, you need to have some sort of some sort of medium, I would call a medium term goal that sounds exciting and different and can get people behind it. And in this case, the major person was me getting behind it.
So then I got more motivated and then like, started like really, we started really flying. So yeah, now the team is back. We're doing like actually really nice, mostly ML focused work. Like we're actually not stuck in like acquiring data land or like the 90 to 80% of the time that a data scientist spends like, not doing it.
We're actually doing mostly ML right now. So that's like pretty cool. And it makes the work much more exciting. Yeah, but that's how you do it. I think so. There's, there's some mix of like, what are your goals? What's the strategy. And then sometimes you need to like, shake it up with some energy, I guess. I wish there was a more precise way to put it, but there's not. Yeah.
[01:00:19] Ken: Well, you know, I think that there's something powerful that like underlying this discussion. I mean, yes, your team was effectively like stripped away from you and the last guy who you'd brought on eventually left, but on the other side of that's an opportunity for you to restart with like your own creative vision in mind and build something that's really cohesive.
Right. And I wouldn't say like, Hey, go fire all your employees to anyone. But the idea of like bringing in fresh talent where that you're hiring, that has something that's more aligned than just like, Hey, we want to work here. And we want to get a wages is a very powerful thing. I also really liked what you did.
I think of it as sort of a paradigm shift around hiring. So for me, I look at that in job applications, right? I've created this ecosystem where I have content out there. I have a lot of presence or social proof. And if I wanted a job, plenty of companies will come to me now. Right. I don't. But there is an option because of the body of work that I've created to have roles that I would probably like based on doing a lot of the work.
I mean, there's a lot of like AI advocates, evangelist type roles that I think would be fun. What I do have now. No, but, but like, those are within my domain and companies reach out. Because of the presence that I've created, right. By putting yourself, your company out there and creating value in this unique space, you've just, as you described, created this pipeline for the types of candidates that you would like to empower to come to you, and, you know, rather than you could do that through recruiters, that's like one thing you could do that from creating job postings, but having a shift away from like, Oh, we have to this, these two ways, this sort of third door, this third channel that you've created is very fascinating to me.
And it's something that we've absolutely hammered home or constantly reiterated with landing your first job. Right. So you described, Oh, I'm going to cold email, right? I'm going to like follow these somewhat like non-traditional paths to advance through my career. Like just because I haven't seen anyone do it or just because I think it might be difficult.
It doesn't mean that we shouldn't pursue an avenue that might be far more fruitful. I mean, the road less traveled in this case, you're going to see a lot of things that other people have not seen both positive and negative.
[01:02:46] Richad: Right. I think if you're thinking principles, one principle I think of is that the more emergent the space, the easier it is to do this sort of thing, to like skip the line to jump ahead to in if that's what you want to do.
A lot of days, I know a lot of people who don't want to lead teams ever, and I think they would make awesome leaders, but they don't want to sound like are fine, but maybe I'll convince them one day. But I think that that's, that's the way, that's the way you need to find like the spaces that. They call it blue oceans. Right. Funny overused business school serve. Right. But I...
[01:03:26] Ken: Porter's five forces. He talks a lot about...
[01:03:30] Richad: I remember, I think that the concept is a good, like you, you can, you can't use it to come up with ideas, but you can look at an idea and be like, maybe this is a blue ocean.
Like after the fact, like post-hoc, you know I think data science is one of those areas right now probably like web three stuff is like that, you know? You could probably advance really quickly because no one knows what the hell is going on. And so you could probably be like, I want to do this. And then people are like, Oh yeah, sure.
Okay. It takes a certain personality perhaps to do that, but I think I almost like that our spaces, the level of structure it has now, as it gets like more structured, it'll be harder. I think to skip over. Maybe software engineers can tell me I'm wrong. And if that is true, I'd love to hear it.
But I feel like in general, it's like the wild west, the more you can build from scratch faster, I think when the space is newer. Well,
[01:04:29] Ken: so that raises a really interesting question. So you've obviously recently brought on new team members, right? And you had to interview them, you had to evaluate and something we highlighted, which was really important was domain expertise.
How do you evaluate domain expertise? How do you evaluate their understanding of, you know, hypothetically real estate or whatever the domain is, and in your unique scenario, how do you quantify if that's important or not? So, as we described, like in a fan company domain, probably not going to be as, as important, maybe you have to have some familiar with product, but in real estate, obviously huge premium on it. So how do you again, evaluate and then determine if it's important or not, maybe it's the other way.
[01:05:13] Richad: I can tell you the fairly strict. It actually fits into the larger structure of how I decided to pick and rate candidates. So I actually made a little spreadsheet for myself and I, and I wrote the job description with this in mind to. Basically when I write a job description and lists like tech and this and that, all these, like...
[01:05:32] Ken: it's important, you wrote the job description yourself too, so.
[01:05:36] Richad: I didn't get to write, I didn't get to write the first part I got to write like afterwards. I was like yeah, the tone noticeably flips like halfway through. I remember being like, can I change this part? We had to keep it, but we could change the order. So anyway so when I write bullets well, first there's the description.
Then I write bullets with every bullet I write down the requirement and the requirement is usually like the general principle, the broader category, and every single requirement has a preferred every single one. And they preferred is closer to our use case, for example, required: Done something in cloud, let's say preferred Azure because we use Azure.
Right. So it's sort of making the case that like, if you don't have Azure, don't not apply. But you do want some cloud where you can like converse with cloud. Talk about it, have used at least like a couple tools in something like AWS, even. Right. And I did that in with real estate knowledge and getting your question.
I think I, as I recall, I put something like a real estate or finance or private equity or I like listed things. And I said like even better, like working with financial data of real estate and something closer to our stuff. Right. So then when interviewing candidates, a lot of them, the vast majority had no domain expertise whatsoever.
This is a thing that probably a lot of data science hiring basically. Whether it's ad tech, which I know very little or many other things, right. So what I did was my interview process did not involve any coding challenge. It didn't, it was no take-home assignment. And there was also no live coding.
What I did was a code review where a person mechanic could share with me if they reached the stage, any piece of code that. After we had already talked about what we do and why, what were vantages and what I'm looking for. And it'd be like, pick the thing you think is most relevant to us. And then just like, explain why you like this code.
And then I'll like ask you the questions. And I made sure to ask questions, like, if it looked like production level stuff, or like a real app, I'd be like, Okay, if you had to make this as a POC, what would you do? And if it was like some little thing, I'd be like, if you had to productionalize this, what would you do?
How would you build it out and stuff? So it was like, my questions were adaptable to what they'd show me. That was one, half the other half was a verbal case. The verbal case was very similar to the problems we're working on now. And so when I would give this verbal case to people, I would be looking for them to ask me the really important.
Preliminary questions like a case interview to understand what's the point? Why do we care? what does it matter? What's the business behind it. And then I was also looking for creativity and breath in brainstorming, like different data sources, because real estate is something that people have physical experience with.
Everyone on earth has experienced with real estate in some sense. So it's in our, in my case, and you wouldn't be able to do this with every type of data in a company's facing, but it would be sort of forcing people to think about their previous experiences with real estate and like logic it out and apply it to like, Oh, I guess I'd want that data because people probably go there at that time of day or something like that.
Right. For example, or I'd want like the street network because of traffic patterns or something, it's like, Oh, cool. Okay. Now you're thinking. Right. So that's like that's so I looked for. Because it was so rare. I couldn't really eliminate people who didn't have real estate experience. If I saw on a resume that they did some like real estate internship somewhere, I was like, cool.
Okay. You're probably in, let's see if like you can use it. But mostly I had to test it in the interview and testing and interview was basically a broader test partly of like real estate knowledge, but more like your ability to think and reason in the moment and have, and have productive discussions that would lead to, you know outcomes.
So in the verbal case, I would go through like that. Oh, maybe what model might you try? Okay. The problem you came up with, then what models would you try to solve that problem? Like, Oh, will we'd predict this. Okay, good. And then I'll be like, Okay, why that model versus this? And which one would you try first?
And stuff like that. It was a very broad, like test of the things that data scientists do. It was also a pretty good facsimile of how we actually work together, which is not like me. Micro-managing you watching you code like in front of you, but more like talking over problems. And then like, now let's go do it, you know, on your own or something.
And it, through doing that, I was able to find like the people who are better communicators, better thinkers from the, from the business side. And we're able to reason about are our use cases.
[01:10:24] Ken: I think it's really important to distinguish how you would interview versus how like a very large tech company would interview, right?
I mean, granted, you are a part of a massive conglomerate, but at the same time, the team that you're working on is focused on is relatively small. You're growing quickly and. These massive companies, right? Let's take Metta, for example, they have a scale problem. So there's so many people applying that the way that they evaluate candidates is going to be significantly more standardized and like, and focused on technical skills.
That's the easiest thing to evaluate, right. At a high level. That is a trade-off for them is they're getting candidates that optimize for certain things. And they kind of plug in as cogs in the wheel in companies like yours and startups. And some of these other companies, you're going to be looking at different types of interviews that are representative of the makeup and the team.
So like you have a little bit more time effectively to sit down and interview these candidates and make sure they're right. You also have significantly less resumes to go through and people applying to these roles. And so as a, as a job applicant flipping the script, like knowing that about the domains, knowing that about the different types of companies.
Ken dictate those types of places you apply to. I mean, I know personally I would absolutely. And I've gone through them. I hate the interview process. It effectively all the Fang companies, like it doesn't optimize for my unique skillset. Right? Like my unique skillset is being able to, I can talk to anyone, right.
I mean, I have showcased that their podcast, right? If someone is a CEO of a company and professional athlete you know, like a, a janitor, it doesn't matter. I'm very comfortable talking to all these people. How can I possibly convey that in an interview at at most of these large companies, I won't make it to that round because I have to do all the technical interviews that I don't have time to study for.
So if you're thinking about, Oh, I need to land this job, like, think about the landscape and the, and the individual positions and like what they're going to value more than just the volume that you're putting out. One thing I just want to touch on is that this kind of min-max aspect of interviewing right there is this idea that, Hey, we want to perform really well in all of our interviews and that's right.
Like I want to do the best possible and put my best foot forward. But if we do in the 70th percentile, 80th percentile on every interview, right, we do well, but not great. You're not going to land any jobs. Right. There's 10 candidates. There's going to be a good person better than you. And the idea is that in a lot of these interviews, you want to actually either do really well, or like, not that.
Because then that makes the process easier for the employer. And it makes it easier for you. If you do absolutely optimally in two interviews or three interviews out of a hundred, rather than doing good in every single interview, out of a hundred, you will have probably two to three more jobs than the other person who does relatively well.
And thinking about the companies and the specifics of them thinking about the interview process and the specifics of them and how you prepare for those things. To me, is something that is fundamentally flawed about how everyone approaches this process, right? Like they're going in with this idea. I'm gonna leak code the hell out of it.
And it's fine if you do that. Right. But you better be the best at leap coding and the best at solving those types of technical problems. Otherwise you're just putting herself in this. Average consideration set. That's probably not going to land you a job. So I, you know, I think your, your story and the way that you're hiring really highlights that right.
Is, Oh, we're indexing really high on this. Like if someone came to you and had three projects in their portfolio, all related to real estate and gave you one of those for that code review portion of the interview, like, unless there are...
[01:14:23] Richad: I never even actually one time this guy, he was looking for an internship. He was in school and he was like, I have a I love real estate. I've analyzed real estate data. And then I used it to invest in a property with my brother. And I was like, what? Really? To this day, the best cold email I've ever had. I was like, look at this another, I was like, I kept trying to figure out how to hire him for like three years.
Like literally, well, two, at least two years, I was like, Okay, our intern, we did intern interviews and all that. And then COVID happened. And then the program was canceled. I was like, Oh, okay. Is it coming back? And like, Hey man. And he had job, he got a job, right. He's out of college. And then I was like, Oh, you know, I have this open position.
I'll maybe you want to apply to it. And then he was like, actually, I feel like I want to do more of this, like the, this investment side rather than XYZ that like your team, like, all right, fine and focus more on the transactions than the operations stuff in real estate. Cool. So that was, that was fine.
I ended up losing him, but that was the, if you, if you meet someone, if someone sends you a cold email that good, then you'll, you'll spend like months trying to like help them. I mean, I have corresponded with long messages, like with advice and stuff to which you seem to appreciate above. So yeah, I mean, it's like. Right. Yeah. Is that it's like the harpoon, you know.
[01:15:49] Ken: Yeah. Well, you became an advocate of that person without with them sending one email and having a couple of conversations. I mean, so I think of it this way. Like a lot of people reach out, they say, Hey, Ken, do you have any open roles at your company? And like, you know, we have five people.
We do not have open roles right now, but you know, we might be hiring an intern or something like that soon, but no one's ever sent me like, Hey, I did this project that's specifically relevant to your company. Or, you know, I've done this project specifically relevant to like your YouTube data or analyzing your podcast or something like you want to talk about something that would open my eyes and like, Oh, you know, maybe if this person did a good job, like I'm always looking for help with videos and, you know, ideas and coding and those types of things like data collection, not fun even for me.
Like if someone wants an opportunity, really not that hard to get one, like a lot of people don't realize how much I wouldn't say power, but if employer or someone in a company really wants to hire someone. They can hire you. They can make a role, they can do this, even in a large bureaucratic organization.
Just like how you got in, you know, that you talked to a bunch of executives in your first data analyst role, they eventually made one for you. Like that happened to me too. I was inter interning before we went to Virginia. I love DraftKings. I was playing it all the time. Doing daily fantasy. I reached out there was an undergrad intern position and it wasn't really related to what I want to do is just like some throwaway internship.
And I applied, I was obviously like way over qualified because I graduated and I was attending a graduate program soon, but I talked with the recruiter and she said, you know, like, obviously you're passionate about this. You want to do something like marketing analytics while. Let's, you know, let me talk.
And they made a role for me. It was a graduate internship, and I got paid like, you know, I was like $5 more per hour than undergraduate internship. And I was like, you know, like it'll, it'll happen If you're meeting the needs.
[01:17:51] Richad: Like, do you know how much thought, how much thought is put into the year's experience bucket on a job description?
I'd never seen anyone think it through. They just put a random ass number, let the dirty secret of the, of the bucket is that the higher you put the number, the less applicants you'll get. And so it just is a way of reducing the volume of things you have to read. That's literally what it is really would be nice is that if we had some way to be like only apply, if you care about our company and industry and stuff.
And like, if there was some way that we could scan like a resume and be like, this person cares. I don't know how do you do that? But I'm just talking out loud. You did. That would be awesome. Right? They'd be like, Oh, this person's like a fit the years experience thing. I'm like, I dunno, like the, I remember where I am now.
Like, I think a job description was written after the fact our advantage and it listed something like 10 years or something on it. I think it's a 10 years. I was like, that's funny. And I'm like, yeah, I don't want to that. I still don't have that. I have like five and a half now, which is funny. Oh, I guess now I could actually apply for those entry level jobs.
But yeah. I mean, a lot of the, a lot of the bullets I'll say like a lot of the job descriptions that you see out there, they're sort of arbitrarily written and like each bullet point they're not necessarily carefully considered. They have a lot of open positions. They're trying to like, they copied probably some other company or person, or like the job description that they put up at their last job or something like that.
There's so many arbitrary reasons. That the job description is the way it is. But if you can, you can skip that if you make a good case that you can, that you care a lot and you actually can solve their specific problems, it's a way to skip the line.
[01:19:40] Ken: Awesome. I love that. I think, you know, it's just, I probably hammered home too much in the podcast, but it's such an important thing.
How are you on time? You, okay. Okay, perfect. You know, since you are coming from the real estate domain, I do want to ask you about sort of the Zillow debacle. That's something that I obviously made a video on. I think it's a really interesting insight into sort of machine learning and AI and data conflicting, or getting caught up in company politics and culture of growth.
And I'm interested in how you sort of perceive. I mean, obviously I was like, Hey, this is a cultural problem. Not a machine learning problem, but in a domain like real estate, is this something that you see fairly commonly? Like what do you, how do you feel about that whole, that whole debacle? If you can talk about it without, you know.
[01:20:34] Richad: Yeah, yeah. I can talk about Zillow. I have thoughts. I haven't actually looked deeply into the company politics side of it. As, as all I have really heard is that they wanted to change. They wanted to make the buying algorithm more aggressive, basically like the buyer not buy, like they wanted to reduce that the bar you had to pass.
Yeah, I guess the most cogent thing I can probably say about it is that I think of when you think of ML products or productizing ML there's an upside and a downside. So the, I remember it it's sort of a running joke, I guess, in all the real estate data people I've ever met that Zillow is estimates were kind of not that good.
Like the yeah, I remember being at CoreLogic and having this discussion with people and we'd find it funny how off we thought they were and the answer. And it, that must be an awful feeling for the data scientists on the other end, who are like working really hard, getting spending years to reduce the error from like 10% to 5% or, you know, and whatever.
And it's probably an accurate because of the complexity of real estate in terms of, in the legal sense, like the different regulations in different places and all the reasons that you can't, it's hard to collect data on. So. When the, what is the downside of a crappy estimate on your listing? The downside is that you laugh at it and ignore it.
Maybe you're still gonna use Zillow. There's very little risk in really in doing that. And maybe if you improve over time, the reputation will improve or something. Right. And you're like, once they got it down below 5%, I believe they're like, Hey, we should turn on. We should use this. And like buy things.
I think that they didn't really consider the fully considered the adverse selection problem. And they also didn't really, I guess, consider how to reduce the risk as they scaled it. So there's a mindset in startups in startup world today, where I guess for the last 15 years, I guess give or take that growth is more important than, than profit.
So they would burn cash, basically ideas to burn cash in order to grow, establish a near monopoly position by destroying your competitors. And then you can crank up the prices and cashflows. I mean, I guess like technology, like the cassette tape and DVDs, and they all go through that sort of like, there's two of them and then they like try to grow and sort of like, maybe we should just grow.
And I feel like maybe Zillow was feeling that, that way of thinking that we should grow, that we buy as much as possible. And then we can have, we'd have an enormous like market share. And then we could do all sorts of things like that to flip and sell whatever. And that's probably part of the mindset that went into it.
See if I were thinking about it, I would be like, if you, even if you did I buy it, you would probably like test it in a very small, like in a small, like considered area that you minimize the potential downside and you demonstrate its benefits. But I guess, I guess that did not happen because of that grow at any cost mindset.
The other main thing I would say is so that's like the that's one end of strategy. No matter what your ML is or how good it is or how bad it is, you need to minimize the risk when you're making bigger decisions from it, right? You see this in self-driving cars too. If someone dies, it's like a news story.
So that's like not good. Yeah. So the other thing is adverse selection. So I see, I saw the stat is either five or 10% of people who were given an offer actually used it. And if you imagine a distribution and you're like, which side of the distribution is taking those offers, it's probably the people with the most asymmetric, with the most asymmetry to gain, you know, if there's an asymmetric game, they get where they have.
They know that no one would ever pay this, but Zillow's paying it. And so they take advantage of it. If you're on the other end of that distribution, where Zillow really low balls, you, no one would take it. And if you're in the middle, you mark, you probably think those are real estate. If you're selling your property, that.
I can probably get more, like I could find some sucker out there, or I don't know, or something like that, but who would, who would buy it instead. But if you see that your, your take rate is really low, that would, that should probably tell you something about any, anything about adverse selection, like which part of it, which side of the distribution, or that is actually taking advantage of these offers?
The third thing, which I found very odd were those viral tweets of people like being able to buy back from Zillow for considerably less than they sold it to them in a very short period of time, which strikes me as like a, one of those sort of smoke tests or data quality checks that you would do.
It'd be like, sort of data quality on top of ML. You would take your output of buying and selling and you would put those same properties info into it. And you'd be like, if less than this is something wrong, you know, it's strikes me as being like, it could, it could have been simple, like it could have been a simple thing.
And they didn't. So it makes me wonder about what processes were underneath. Probably there was really great ML. It was the metal stuff around the ML that ultimately led it to fail as a business model.
[01:26:00] Ken: Yeah. Well, you know, it's interesting to me. So the CEO of Zillow, he comes from like a pure transactions, fast online transactions background.
I think he started the blast door and a couple of other companies that are like, boom, boom, boom, fast. Real estate is not one of those markets. Right. And to me it's interesting because the volume is not that high. You know, they, they did like less than 10,000, right. Property sale purchases, but the volume is low enough that they could have had a human ML hybrid model very easily.
Right. They could have started it that way and improved it over time. And also like the data quality probably isn't high enough where it would make sense to do something that's almost completely autonomous. And I just don't, I cannot understand why they would approach that problem in that way. Like this to me is a very hands-on problem that you would, that you described.
There's like human errors, right? Like there's people that are taking these deals are finding opportunities and data effectively only use it to buy through, you know, to get arbitrage on their part. And I just don't I'm, I'm disappointed that like the need for a low-tech counterpart to a technology option.
Is the downfall of a really interesting technology solution.
[01:27:26] Richad: A big damn shame.
[01:27:28] Ken: Yeah, well, but it's also a, hopefully a learning lesson for that. I mean, or for any other real estate organization in that space, like Redfin, I think is actually killing it, doing something very similar, but they just have Redfin agents and they've absorbed that process and are really set up well to do something very similar.
But with these handholding, the handholding that didn't need. So Zillow's loss to me, it looks like Redfin's when,
[01:27:54] Richad: It could be in my last place before Ivantage was that gray stone, which is a, as far as I thought, there still are probably still the largest lender. Multi-family lender in the U S they basically held the gate, Fannie Freddie and HUD loans.
And we were part of, sort of a lab, like their sort of little tech bubble to innovate how that was done. Like with more technology and less. Cold calls, you know, and like looking in databases and scrolling manually and stuff. So that was, those were interesting data data problems. Cause you're trying to make some decision related to like, what loan should we give?
What cap rate what's the size of the property, that sort of thing. And how do you do that at scale? It's a pretty tough one because you have to bring together a lot of different data sources. So I feel, I feel for those who are trying to like make real decisions from those real decisions involving hundreds of thousands and millions of dollars, you know?
Yeah. Well, yeah, you need to minimize the downside risk through like tests through like testing a mindset of a scientific mindset, a mindset of let's is this true? Let's test it. Even if the simulation works, let's see if it, how it works in the real world. And let's, what's the, what's the maximum downside risk of doing this.
Let's like, and then slowly build that up as he proved that it works. I see, it's the only way to do it in something like old, like real estate. Cause you, you get that and then I'm sure there's thousands of real estate professionals out there, not in data who look at Zillow and they're like, yeah, that AI stuff is dumb. Like we'll never do it, you know, or something.
[01:29:32] Ken: Oh, it's funny. I mean, you look at it like the short term, like costs and ROI doesn't make sense, right? Where I could hire three real estate professionals who are really good, the best in their field for what hundreds and hundreds of times less to do the 10,000 deals than it takes to develop a model.
That does a worst job. And I mean, at the very least you do do the hybrid because the cost of hiring professionals to evaluate it is so small compared to the production cost of the model. I just think it's a very, like, from a data perspective, from like a business perspective, a very puzzling problem.
I know like the guy who started deal is done like four or five companies already and is a good Jillian air, but it's also like he should be business savvy enough to avoid a pitfall like that, which, which was baffling to me. So those are, those are all the questions I had. I'd love to hear about what you're working on now. How people can, can get in contact with you and learn more about you as well.
[01:30:37] Richad: Yeah, sure. Well, mostly we'll see on LinkedIn at the moment. So I'm working on two side projects at the moment. Related to this story earlier of how I turn around the team and data science here. I'm writing a report and interviewing data science leaders, current leaders, aspiring data science.
They do their new wants. Who's interested in leadership stuff. I'm basically interviewing them, one-on-one to learn what tips they would share, what they wished they knew when they started in data science or what they don't know about being a leader that they really want to know. And I want to write basically the thing I wish I had when I started, I wanna, I wanna write something for me from three years ago that I could have given me from three years ago.
If relativity didn't work the way it does, you know, we have the technology, I'd be like, read this don't don't do these pitfalls. Make sure you've defined ROI early on, you know relationships are great, but make sure you also use like focus on these specific numbers and like finding that quicker, you know, among other things, right?
Like how do you distribute the work of a team and how do you start from scratch, like a more efficient way to start from scratch, which also by the way is a very interesting ML problem, like a warm start or how do you like initiate an algorithm or something? You know, that's like a big problem in in, machine learning.
It's also a big problem in leadership and in building a team. So I think there's a lot of parallels that I'd like to explore there. That's one thing I'm working on. The other thing is I realized that there's a big gap. The world of data science in terms of like I have inside sales skills that you, you know, you can use to basically skip the resume drop that I did especially good.
if you don't have the degree in it or you're mostly self-taught and, you know, while you work on your side projects and stuff, like how do you, how can you have productive conversations and grow your network in a really effective way? Building something, probably gonna have an event in a couple of weeks to like, teach people how to do that in a more like focused and in-depth manner.
If you're understanding either of these things, I would just tell people, like, reach out to me on LinkedIn, like connect with me and then mention which of these things you're interested in. And then yeah. 'cause, I'm not, I'm not so big. Like you can yet where I can be like, Oh, there's too much too many people.
[01:33:12] Ken: Well, I'll make sure to add a link in the description to your LinkedIn. And you should just like underneath your, your face there have here, your LinkedIn popping up for most of the episodes. So everyone who wants to learn more, feel free to reach out to Richad. As he mentioned, he really does enjoy the mentorship stuff.
So I will forward all of our requests over to you now. Just kidding.
[01:33:35] Richad: That'll be 20 bucks.
[01:33:38] Ken: This was so much fun, Richad. Thank you so much for coming on. Really enjoyed it and until next time.
[01:33:44] Richad: Awesome. Thanks, Ken.