Why Attend Data Science Conferences? (Sheamus McGovern) - KNN Ep. 92
Updated: Jun 28, 2022
Today I had the pleasure of interviewing Seamus McGovern. Seamus is the founder of the Open Data Science Conference otherwise known as ODSC. One of the world's leading successful AI conferences. ODSC runs numerous virtual and in-person data science conferences around the globe. Seamus also founded AI plus training, which is dedicated to bridging the AI talent gap with on-demand live training programs. He's a technologist with over 25 years of experience in software, data engineering, and AI. And in our conversation, we dive into how Seamus integrates project management into his life, and how his passion for learning was multiplied when he first found conferences. I happen to be speaking at one of the ODSC events coming up, and I hope you'll join me I'll be there in person. I'd love to meet all of you. I really enjoyed this conversation with Seamus he's fun, he's animated and I think you will too.
Open Data Science website: https://odsc.com/boston/
[00:00:00] Sheamus: You know what I mean? And that's a key part of work-life pass because if your work-life is sucking the life out of you, but that's affecting your home life, right, that's affecting your personal life. And if you're working with people that you love working with, you love collaborating, when you love creating what it doesn't matter what field you're in.
[00:00:24] Ken: Today, I had the pleasure of interviewing Sheamus McGovern. Sheamus is the founder of the Open Data Science Conference, otherwise known as ODSC. One of the world's leading successful AI conferences. ODSC runs numerous virtual and in-person data science conferences around the globe. Sheamus also founded AI plus training, which is dedicated to bridging the AI talent gap with on-demand live training programs.
He's a technologist with over 25 years of experience in software, data engineering, and AI. And in our conversation, we dive into how Sheamus integrates project management into his life. And how his passion for learning was multiplied when he first found conferences, I happened to be speaking at one of the ODSC events coming up and I hope you'll join me.
I'll be there in person. I'd love to meet all of you. I really enjoyed this conversation with Sheamus. He's fun, he's animated. And I think you will too. Sheamus, thank you so much for coming on the Ken's Nearest Neighbors Podcast. So you're obviously behind some of the incredible conferences at ODSC. You've had such an interesting career.
You also have a really excellent sounding accent that your accent is probably one of my favorite sounding ones of all time. And I'm happy to have you on the show to tell your story. To have some people learn about, about your progress and what you've done in this space and how you've been able to build these incredible community. So again, thank you so much for coming on.
[00:01:48] Sheamus: Ken, awesome to be here. And you know, I can only reciprocate you were on my show, so it's good to have you good for me to be on yours.
[00:01:55] Ken: Excellent. I had a blast talking to you. We had a really good chat over LinkedIn Live, and I hope it's recorded somewhere. I'd love to be able to share that around us.
So perfect. So the first question that I always ask to sort of get people introduced to you and your story is how did you first get interested in data? Was there a pivotal moment? Was something, was it something that happened like you had this class and it just blew your mind? Or was it a slow progression over time.
[00:02:22] Sheamus: Yeah. Yeah. It was a pivotal moment of frustration, I think is the best way of describing it. So my background is in finance and software engineering, and maybe we'll get into that a bit later, but I was part of a startup here in Boston. I just returned from London and doing a capital markets startup in the bond markets, which sounds super boring and it was super boring.
But the crux of the matter was we were trying to mind all this information what's going on in the stock market, going on in the bond markets and you know, being a data engineer knowing quite a bit about statistics and other stuff like that. I thought I had a good home lessons were struggling to certain pieces.
So I went to a meetup in Boston and after attending a ton of conferences, I was like, couldn't believe here, I'm at a meet up. It's free. And it was a, I think it was a startup was called Hopper. They were applying NLP to travel arranged, you know, travel booking and stuff like that. So after a talk, I met with the speaker and you give me massive insights into how to use NLP to model data and scrape it.
And that's really where I kind of felt that I need to migrate from this idea of being a software engineer and into the realm of data science and machine learning, because this thing is incredibly powerful and such a different way of even thinking about projects than we used to in software engineering.
So that was pretty exciting. And that was a big moment. And then a whole chain of events happened from there, which we'll get into later, but that was the beginning of it.
[00:03:47] Ken: I love that. You know, there's something, I think that there is sort of a paradigm shift or a mindset shift when you're coming from a lot of other disciplines into data.
So the first thing that I noticed I had a lot of physicists that I work with or, or physicists that worked for me as data scientists and from coming from physics, coming from a lot of other diseases. There's a formulaic approach. There's a right answer. There's no estimation like, Hey, we're trying to model what's happening in the world.
And we do that with an equation. And to a certain extent, I think you do that with software engineering too. There's like a binary right. Or wrong. And in data science there that there's this idea where like we're working in shades of correctness, you know, no model is a hundred percent, right. We're just trying to create a do better than what exists.
And that idea of that incremental improvement, you know, as we've talked that sort of pervades through your entire life. And I really, I really liked that, you know, how do we do just a little bit better? How do we iterate on this and tinker and improve. And I'd love to hear where that mindset developed. Was that something that you got you know, earlier in your career when you were working in software event or is that something that you've developed over time.
[00:04:58] Sheamus: Yeah, that's a super interesting question. I think it's a discipline that comes with time because if you're in the software engineering field, the engineering field in general, like you do adhere to certain, you know, project management principles, but then you might use a methodology such as the Agile methodology.
So, you know, starting, I think in the nineties, this whole idea of doing these quick sprints, quick turnarounds, a set of deans doing these monolith three-month projects. Cause when I started my career, you could be on a project for three or six months. That's incredibly boring. Cause he would spend the first month in doing system analysis system requirements and then you move into the bill phase and then the deploy phase nowadays things are much faster.
You know, you're very much focused on a a one week or a two week sprint. And after a while you start to from doing that from a few years, you start to like apply that to other aspects of your life, right? So you're used to understanding. How do I get quick wins? Right. You can look at life and everything is, you know, whether it's your personal life or what is your career, your career.
Everything is kind of a bit of an opportunity to concentrate. So being able to the whole concept of being able to work on something for a week or a day or an hour and deliver and beautiful results in a quick turnaround time is pretty powerful. And I think that's very true of whether it's coding on the job, learning on the job are just training yourself.
Like it's really kind of hard to say, Okay, I'm going to become a programmer which could take five years, or I'm going to become a data scientist. Like how do you break these problems down into smaller manageable pieces? So it doesn't matter if you're tackling an engineering project, are you're an analyst or a business analyst and you just want to tackle the product.
So that whole discipline of breaking things down and applying that discipline to it, I think has been massively powerful. I think a lot of people have gotten good use for that.
[00:06:51] Ken: Incredible. And so I think that there's a really cool sort of gray area between. Project management in your work and project management in your life.
I mean, you've described to me that you basically run your life with the project management framework. I wanted to talk about this later, but I w while we're on the topic, can you talk a little bit about what your personal project management philosophy? So what's your like Sheamus management philosophy?
[00:07:16] Sheamus: Well, this is a bit like ... like, what's, these is the way to think about this. So, you know, I did study project management first and foremost, cause I was pretty fascinated about that because you know what I knew to be successful, you had to build a good team. So how do you do that? And the project management people think it's all about, you know, getting stuff delivered on time, but projects can fail for a lot of reasons that can fail because let's say you didn't bring it in on budget.
For a work project, they could fail because you know, you didn't deliver the features you promised. They can fail because you didn't calculate all the risks. So project management, as they say is a as essentially a study, it's a risk study, right? It's an exercise in risk and Like many things you can strap, let that into opportunity cost.
Right? So it doesn't matter if in your work life, your personal life, everything you do has got an opportunity cost for, right? You decided to go and watch the matrix, reloaded the revolution, you know, then you're giving up in some other movie you want to see, right? Every, every decision you make if you go eat at one restaurant, you're excluding somebody else.
So it doesn't matter if your project you're working on your career or your personal life. I think understanding opportunity costs and putting a value on that is a great way of looking at your kind of work-life balance. Like where do you want to put stuff? Where do you want to put time in? So I think, you know, in short that's where I think of it, a loss like in terms of what's the cost of this to me.
So it doesn't matter if I'm spending five minutes on something at five hours or five days, I am weighing it against the risks and the opportunity costs against something else I could be doing. I think that's a very important.
[00:09:00] Ken: I agree. I, so I, one of the reasons why I didn't pursue playing golf, something that I loved further is that I realized there's an opportunity cost of me trying to play professionally over a period of time.
Like, you know, my job opportunities that I have potential access to dramatically decrease. The longer I played golf, the amount of learning that I could do, the amount of time that I could spend doing other things, learning and educating myself one way down, something I think is really important to sort of extrapolate on associated with that is when we talk about cost, how do you view costs?
I mean, is it in dollars and cents? Is it an enjoyment? Is it in you know, fulfillment? Like what do you, what is your like major KPI cost cost a metric that you focus on in care most.
[00:09:51] Sheamus: Yeah, I will say as a quick aside, can I bet you're still a better golfer than I am despite putting that much time. And I'm an Irish man. I should be a good golfer. I always think of in terms of like, you know, everyone, everyone has their own, you know, their own environment, their own family cost to me is. When you're, you know, I really do try and strive for that work-life balance first and foremost. Right. I kind of divide my days into, you know, I've got my Monday through Friday and then I try and spend at least half my weekend into Something at work and then the rest is kind of family time, but just walking, talk to me about the work side of it.
Think about any project you want to undertake. The problem is, you know, where we want to be builders. So you tend to just dive straight in, right? You're just diving straight into a project and from a data science mindset, it's a very research focused discipline, and that's a bit of a trap because I've seen data scientists quickly go to, Okay, you know, I'm going to do a pilot project here.
I'm going to spend three or four days just gathering some data and see what I can find out about it and kind of taking a bit of an undisciplined approach. And that's why a lot of data science projects failed because people think in the underestimate, the amount of cleaning, the amount of fan modeling, the amount of feature engineering you have to do as well.
So one approach I always do is kind of do, and even as a back of the envelope thing, like, Why am I undertaking this? What's the outcome? So starting with outcomes, starting with some. Goals first. Right. And again, I'm not saying like, put this in the formula or something like that. It's just that if I'm going to like, let's say for ODSC, we're trying to figure out, like, I'm very passionate about a topic lab called machine learning engineering.
And I'm talking to speaker committee about machine learning safety you know, autonomous vehicles crashing and all that stuff is a word pursuing that. I do a little bit of initial research to see, you know, what's going on out there and stuff like that before I would start getting other people involved.
And it was scraping, I would try and say, Well, really, why do I want to do this is because I want to do what ours is going to have value in my community. So sometimes I want to do for the sake of doing myself, which can be an awesome. Other times like this is really for my work or for ODSC or something like that.
And as I said, as a, for a really a personal goal are as a, for the outcomes, I think. So I think being honest with yourself as to why you want to pursue certain things is is interesting. And it's also an interesting discipline because I'm surprised about how few people do that. And I'm the same way when people ask me to do stuff.
So you know, basically self-employed, but I work with a team, we all collaborate together and if I'm going to ask them to do something for me, I got to be damn sure that I haven't got to use what they, what they do for me. And I kind of applied the same discipline. So when people ask me if stuff, right?
So when Ken Jee asked me to do an interview on the show and I said, Ken, you're awesome. Of course I will. And I would love to talk to your audience. So, you know, you do that quick mental math, but it's surprising to me how often we commit to doing stuff subconsciously. And then all of a sudden we feel it. Right. So prioritizing, prioritizing is a key part of that. Yeah.
[00:13:06] Ken: I really like that. I think my view, there was a little too reductionist. I mean, what you just just described is that a lot of different things, there's different reasons for doing them there isn't, you know, I was taking that purely math mindset.
Everything is broken down into ... and I think in life, that is, you know, we can't have those oversimplifications. So, you know, I agree. I think that there's something really important about scoping and estimating correctly and evaluating the value to you upfront something that I have always been trying to get better at, and something I personally struggle with is properly estimating my own time and improving it, that meta skill. How have you been able to develop that or, you know, like how do you, is it just reps or is there, just, you know.
[00:13:56] Sheamus: Yeah. Reps, reps, reps. There's there's only so many times you can count to 12 in your life. Right? My trainer told me that once and I was like, I was stunned. I was silent for five minutes, which is an achievement in itself. I think how have I been able to achieve that? I think, you know it's been very important. Sorry. Can you repeat the question? Cause I just had a COVID cloud.
[00:14:22] Ken: So, I think the question is how do you get better at estimating your work or how long things will take or evaluating things along those lines?
I think that's something a lot of data scientists struggle with is that, Oh, I'm doing this EDA on this new data. I can't possibly estimate that because I haven't seen the data yet. I haven't done X, Y, Z yet. And that's sort of this pitfall that a lot of people fall into. Whereas I think I've seen some data scientists, like, yeah, it'll take me roughly X, Y, Z time, and they nail it every single time. And it gets me so frustrated because I struggle with that so much.
[00:14:57] Sheamus: Yeah. Yeah. So, so that's an interesting question, because I think to make the obvious answer, it depends on experience, right? So again, I'm a big fan of the Agile method. And if you're people are listening to this, like I think a lot of times we focus on, you know, data science as a discipline in and of itself.
So if you're a beginning data scientist, listen to his podcast, I would definitely encourage you study agile method, right. And study what a Scrum Master does. It literally takes you an hour or two. So I love the scrub discipline of crowdsourcing estimation. Right? So you get with your colleagues, you get with the team and they tell you from that aspect.
And that starts to give you a sense because it's like, you've got a friend who's always late. They will think they're always on time. Right. So that will tell you if you're an overestimated. An underestimate or a spot on, right? So getting feedback from other people and understanding where you kind of fit within this thing is very, very important.
So I think the rule wasn't number one, there's no dye self, right? And you're not going to know yourself. And that sort of people are telling me that. And I've seen even outside of Scrum that discipline, you can actually do this with your boss. You call it, the boss comes back and tells you how long it's going to do this.
You can kind of make this part of a conversation because especially in your career under pressure, like you're thinking yourself, well, I don't know how long it's going to take. It's gonna take three days. It's going to take five days. So first of all, giving yourself a range is good, but I'm having a conversation with someone who's asking you something about how long it's gonna take.
It's gonna take three or five days. I'm not saying you say that your boss is like I Sheamus, how long will it take? Well, John, how long do you think it's going to take? That's not a good, that's not a good career strategy. So, you know, those were some very, very basic trip tricks there. And then it always comes down to, you know, I know people overuse this, but it's known knowns and known unknowns, right?
In software engineering, one of the reasons I stopped doing it, we got boring. Like I pretty much could tell this product is going to take X number of backend engineering, X, number of software architecture, or X number of front-end and all that kind of stuff. Data science is a much, much different discipline.
So what we tend to do there is instead of like making a big bet, this is going to take six months. We used to say, I'm going to take one week to do did exploration, data profiling, things like that. And then I'll deliver next as of product. So I think another tip for your audience is not to get hemmed into a commitment that you've no idea you're making.
Right. Like let's say, like, let's say you're working as a consultant where they want an award, a contract for you. That's even more difficult than the boss. So understanding that sound. There's research aspects to data science, which has lots of unknowns. That is a research project, not a deliverable project.
So that's a key part of understanding that aspect, like someone may say to you, I want you to build a data science app and, you know, you may know the software engineering part pretty well, but you can still say, well, that's a research part and that will require a certain amount of time to do that. So for the last 10 years I've been doing a lot of consulting and usually, yeah.
People are always pressuring you, like, tell me how much this is going to cost you. And how much time is it? A two month project, three month, six month project. When I tell them is I said, give me a small amount of money or don't give me any money and I'll spend one or two weeks and I'll probably spend save you a half million dollars because by doing one or two weeks of research, you can save a lot of time.
And. You know, make sure that you don't set up a project for failure. And we've done that multiple, multiple times. And it still amazes me to this day that instead of spending $2,000, or $3,000, or $5,000 for a week or two, not trying to give away my rates or anything for a project that could cost like hundreds of thousands of dollars.
They're like, no, we just, we just need a time estimate. Cause we've got a bunch of stuff like that. So people are not good at, they're not trained to do this and they're not good at doing it because they're asked to do reasonable making reasonable estimations. And that can be a trap. You should be very, very wary, especially in your area career.
Right? Because once that word comes out of your mouth, like I can do this in two months and then you're gonna be in misery for next six months while you try and please the damn project. And yeah. So.
[00:19:14] Ken: I love that. Well, I think two things come to mind when you bring that up. And the first is the, I think it's the Parkinson's principle.
So it's generally the task that you have expands to fill the amount of time that you give it. And I've seen, I had sort of this light bulb moment and you were talking there where it's like, well, if I say that I'm going to do two weeks of exploratory analysis, right? Like I can do just two weeks of analysis.
If I know that that's a constraint, I'll get X amount done. If I had three weeks, maybe I could do more. But, you know, have you ever seen those? Maybe it's because I watched YouTube, but there's a video where an artist though, they have 10 seconds to draw a picture. They have like a minute and then they have 10 minutes.
They draw the same picture and you get like the varying like level of detail in whatever it is. And to me, it's the same thing with a lot of this work is that, you know, if you spend less time, the quality is going to be a little bit worse, but you're still getting some sense, some feeling of what it is. And you can always expand it further if you want later.
[00:20:17] Sheamus: But I'll tell you a funny story about that. Cause I had a manager ask me for that once consulting, kickin, finance, and the guy's like, well, I want to got you only got a day or two. So figuring out what you do in a day, I came back with a linear chart.
He was like, what the hell kind of a model is that? I said, it's a linear model. This is what he wants something in a day. I didn't have exactly time to train a data set and building your math. So there you go. So it's kind of like your, your one hit of sketch there.
[00:20:47] Ken: A real-world example.
[00:20:49] Sheamus: When all else fails them. Linear algebra is your friend. Yeah. Yeah. Simple, long chart, but anyway, sorry, I have to dress, but I couldn't help it.
[00:20:58] Ken: No. So I think what you're describing also with scoping process.
[00:21:01] Sheamus: Ken, if you were listening, it was a candidate, another candidate that gave that to me. So Ken, if you're out there listening, that was you, I'm talking about.
[00:21:08] Ken: I love of that. So, well, something that you described with the challenges of project estimation, that's something I see at a macro level with large companies too. Right? I think a lot of organizations, they spend a lot of money building out data science teams, and then they didn't have, they didn't know what the data science teams were supposed to be to do.
And so they treated them like other teams, like in a year, we're expecting these types of results. And I would say that probably the majority of large organizations that just hire data science teams without a roadmap, those teams significantly underperformed, largely because of the scoping and the planning, because they were looking for ROI when they had to do a lot of preliminary research to even test to see if any of the projects that we're working on were worthwhile. So it's funny how something at a micro level, if we're talking about really expanding is it is highly visible at the highest level in these big organizations. And it's a little scary and I kind of, I wouldn't say sad, but it's interesting to see how much that specific problem has cost a lot of companies.
[00:22:20] Sheamus: Yeah. And there are some very high profile examples of that. And even, even in the last couple of months, but. But I will tell your audience if they're listening. If you're thinking about becoming a data scientist, machine learning engineer and you hate hate math, and you're a crap coder and you just like staring at a computer and coding is just pure torture, but you want to be a data scientist.
I've got a much easier route for you. Just go study project management and be a data science citizen, be a data science enthusiast, but there is a massive need for AI project managers, machine learning project managers data science project managers. I don't care what you call them, but it's exactly that.
At the end of the day, software engineering is about building rules-based systems and rules based systems are very predictable. You get her a set of requirements, which translate a set of rules, and then you build your platform around that rules. So that's very easy to project manage. And then you want to, you want to add new features.
You have a new set of rules with data science. You exactly said that th that discipline of being a project manager, being a data science manager is much, much different, because think about all the stakeholders you have to involve, you could be dealing with the C-suite. You could be dealing with the product team.
You could be dealing with the engineering team, the data teams. They need to understand the different set of criteria around a data science and machinery project that requires a certain amount of research, a certain amount of unknowns. And I've even seen counties. Who've done a well, they may start off as a research, a low stakes research.
Then all of a sudden, because the research project was successful, they think now we're going to commit to let's say, just give to protect the guilty. We'll give just a vague example. So let's say you're a banking company and I've actually done a project like this. You're working on a new credit card approval system and you're running some models and trying to do what and improve certain things.
Since the pilot project works. But it's not real, it's not it's pilot project for our limited number of uses. And then all of a sudden like, Hey, we did it. We know how to build this, we've done it. Now, we're going to build a production version of this like production version is much, much different than, you know kind of research type project because now you've got a much stronger discipline applied to it.
There is so much larger, right? You're making commitments, you're bringing in cross disciplinary teams who may not be familiar with this with data science and machine learning and the fuzziness that goes on around it. So I think that's a great career choice for a lot of people. That's kind of getting into imagine role in data science and AI without necessarily being a technical data science or AI manager himself, AI professional themselves. So it's actually a big problem which presents a lot of opportunity.
[00:25:07] Ken: I love that, you know, someone I had on the show before Greg Coquillo, he does a role like that at Amazon. And he gets to work on all of these cool things and he doesn't have to have the headache of writing a lot of the code. I actually even took a project management role and it was more, it's like a, it's called a business consultant in analytics.
So I'd interface between the business stakeholders and the teams without having to directly manage the projects. And then eventually I moved into a product owner role, which is the, you know, the best of both worlds. You get to effectively see the results of a lot of these projects. We, in these conversations around the strategy around them, but you don't always necessarily have to do the hands-on work.
I will say actually like the hands-on work. And that's why I stopped, right. Is because sometimes dealing with people as much as I like talking with people, I don't like necessarily telling them what to do and disagreeing with them all the time. This is a lot more amicable I got to talk about and enjoy the process of talking to other people rather than it being as combative.
[00:26:09] Sheamus: Yeah, but I will tell your listeners. What's fascinating though, is as you kind of alluded to there, like there's, there's so much going on in the space right now, and people, again, keep forcing on, you know, I'm going to use TensorFlow, PyTorch, I'm going to use Python, but think about all the things happening around responsible AI, right?
Privacy trust and explainability, observability, right, and then something think about something that we're focusing a lot on ODSC, like machine learning safety, I'm kind of fascinated by this subject. Maybe we're too early, but you're looking at things. Who's got control over autonomy systems.
Whereas civilian military are whatever role how do we run models that are ensure we got privacy across multiple systems? Like so many. So many projects are using data scraping the web. How do we know there's not poisoning attacks in those, in those data sources? So what's fascinating about these new roles is if you're not necessarily hands-on you get to deal with all these.
Fascinating concepts that people didn't even come along two years ago. If someone told me like five years ago, Oh yeah. The way they're going to be able to PIs in your data science machine learning projects with data that they placed and that they put in place in the web, like two years ago, I would have been like, what are you talking about?
So some of the stuff coming down the pike is fascinating. So when you're dealing with a business, like you gotta be very cognitive stuff, you know, your business, male. Yeah. Bias and AI, facial recognition, you know, like you pretty much got assume all your data you're collecting is bias, you know? Cause there's no such thing as the personal universal set of data.
Right. You gotta like, how do you explain explainability observability to your business? How do you explain model risks? What can go wrong? What are the known risks and your known risks? So, you know, the engineers is kind of like in software engineering, you have. The developers, you know, they, they, he, she, they give birth to this beautiful thing, this, a software program.
Then you have these QA people come in and try kill it. Right. So we need more of that. And then the science itself, you know, we need people, you know hacking at these models, poking holes in them, seeing all the ways they can break. And they're so much tougher to break than traditional software, but they do.
Right. So that's, that to me is why I kind of get passionate about this because there's so much coming down. Like it's the universe of opportunity and challenges within data science and AI. It's like the universe we're occupying right now. It just keeps expanding at a rapid pace.
[00:28:42] Ken: This episode of Ken's Nearest Neighbors is brought to you by Z by HP. HP's high compute workstation-grade line of products and solutions. Z is specifically made for high-performance data science solutions.
And I personally use the Z Book Studio and the Z4 Workstation. I really love that the Z line who comes standard with Linux, and they also can be configured with the data science software stack. With this software stack, you can get right into the work of doing data science on day 1 without the overhead of having to completely reconfigure your new machine.
Now back to our show. Well, that's such an important point is that, you know, these data science. The systems and the guests, the checks are not as readily in place as a lot of the software products that go out. I mean, I think about some of the work that I've done. And in some of those times I was the only data science working data scientists working on the project, right.
There was maybe people were doing code review, but there wasn't a second set of eyes necessarily making sure that the assumptions of my model were correct, making sure that you know, those assumptions on the data were correct a lot of these types of things. And to me, that's a really necessary step as we move more towards these massive production models that have massive implications, as well as, you know, even if it's a third party team or another team within the organization, just bedding a lot of information or models that go out.
That's a really. Valuable thing that I'm not seeing done at scale as much as I think it should be. I mean, I think a lot of the problem is that in the largest organizations, that absolutely is, you know, the really tech forward companies. But if you look at it, the majority of data science is probably not being done at those organizations.
Right. There's only so many. And so we're probably missing a lot of oversight with a lot of models that go out because you know, that's just not standard practice and every organization today.
[00:30:41] Sheamus: Yeah. So I agree a hundred percent, but I will say there's a tidal wave of solutions coming for that. And pretty much everyone's putting it under the ML Ops.
Umbrella all opposite is a big tent, right? And so ML Ops, machine learning workflow, anywhere from your data generation data capture all the way to, you know, continuous integration, continuous deployment of model to looking at model desecration adversarial attacks and models of that. So a lot of stuff was falling under ML Ops, but I think the industry is making a slight mistake there.
Like, you know, if you look at the number of startups right now, I think. 50% of startups in the AI ML space to come out in the last couple years are actually in the envelopes. Well, I disagree. I think 25% were in DevOps space. Nearly 25% are pivoting and was like, I don't know any company out there just won't be on with us.
So that's good news. Right. But they're all focusing from a workflow standpoint and then trying to throw in ML safety machine, like machinery said 200 out. I predict machine learning safety is going to be as big as envelopes in a couple of years because that's, again, they're there too. They all have the integrators of course, but the two two, two hugely important topics and they're both big umbrella things.
'cause we just did a blog for our blog site to today where myself and Alex wrote this blog on you know, responsibility, I tool kits and open source, right? There's the IBM 360 tool kit there's TensorFlow has got some bolt onto their platform as well, but it kind of see like the open source community is slowly waking up to this because, you know, you look at number of stars, number of downloads.
It's not that great, you know shopping live without, for a long time explainability number severability, but you know, it doesn't have the same kind of numbers like you know, TensorFlow does, so the community starting to slowly embrace it, but I think there's going to be, that's going to be a hockey stick growth because, you know, I still think A lot of companies are still in the exploratory stage or they're using very simple well-known models.
Right? Basic, random forest logistical regression. Those models were pretty good. If you're Zillow you're, maybe they're not working so good, but sorry, I just couldn't put that.
[00:33:06] Ken: That's all right. I'm probably never getting a job there after the video I've made, so I'm not worried about it. So, you know, I think it's really interesting how that's integrated and the adoption over time.
So to me, a lot of these things are technology solutions to human problems, right? Like, ethics is a human problem, not technically a technology problem. Right? What TensorFlow and a lot of these other packages that are doing they're solving technology problems, right. And bridging that gap between technology solutions to human problems from technology solutions to technical problems is something that I think does take time and there's lag and there's lack of understanding.
But I also think that the, like the solution to any of the issues that you described are going to be more technology solutions, right? It's a Pandora's box thing, right? We've already unleashed this technology or not putting it back in the box. What we're going to have to do is create even better technology to monitor it or to make sure it's more ethical by our standards and things along those lines.
So it's sort of a really interesting Problem that we see in that. And I'm excited to see what comes out of it. I know it's a little bit outside of my tactical understanding a lot of the time, but it's fascinating, at least for me to read about and try to try to conceptualize.
[00:34:30] Sheamus: Well, I will make a bullet prediction here. I think I've only done this once or twice before, so people can hold me to switch, but I am when stuff blows up. Metaphorically speaking, of course, we still have a close up on only dad do people pay attention and you know, you can see what my gray hair is. I've aged my, maybe myself a bit here, but you know, you experienced like this pandemic we're going through right now.
This is like a monumental experience for a lot of people. The, prior to that, the the great recession as they call it in the us are the collapse of the financial markets, right. That was a huge deal. the.com bust all of these are kind of seminal moments in technology and societies cause right.
So when a financial markets collapsed they realized quite quickly they didn't know what the hell was going on in the financial markets. They had a regular. They had a control system around it, but didn't really know what's going on. And it's kind of funny what you're saying about, you know, a lot of reasons these models blew up is everyone's chasing the same model.
Like a, you got a couple of PSTs working in fashion markets. PST is very hard to come by group think. They thought they were modeling the right things. And if you look at the history of the financial markets implosion around you know, trenching bonds and stuff like that, that was based off one guy's model.
And they only used 10 years worth of. And based on his model, the the stock markets the housing market never collapsed and absolutely wasn't his fault, but people took his model for batim. So my prediction is it's not just a technology problem, it's a societal problem, and it's definitely a business problem.
So if you look at regulated businesses, take finance and health care, for example, they have a whole industry built around compliance, right? Even construction, you have to have compliance. So buildings, don't collapse and people don't tie. I think people have been predicting this for years and it still hasn't happened, but there needs to be more of an understanding of what's going on in the industry, because look at what's going on with, I'm not gonna name names, but look, what's going on in big tech.
Right. And the issues and to law suits that keep coming out of there. At some point. We need to understand that there is a lot of risk involved in this and you need to build system companies. They need to build systems similar to compliance, right? So no matter, no matter what kind of rules you do as a government, our society around this stuff, companies really have to be at the forefront of doing that.
Right. So I think that's going to be a big business for a lot of companies. I think it's going to be a great career career opportunity for a lot of companies. And then when that opportunity comes, it's going to come with. Right. So when a financial markets collapsed there was an obscure job called Chief Risk Officer.
No one cared about you and no one talked to you that it talked to you, they would listen to you. And then that became the hottest job in finance by far. So I think you're going to see something similar in the AI space and that hasn't come yet because I don't know, you can read all the polls and serve as you want, but I would guess, like only of the companies that say they're doing AI would say only about 25% are really applying, you know, kind of cutting edge modules as opposed to very simple models.
And the rest are keeping it in a safe space, a safe bubble. Right. They're not willing to apply core business strategy to AI yet. Right. So, you know, Yeah, look at Netflix and Amazon, right. There are businesses built on a recommendation model, but it's still not fundamental. If their recommendation model collapsed tomorrow, they'd still be in business.
You know what I mean? Netflix recommendation model to me is like, what the heck? I'm just, they're the best model they have is like what everyone else is watching, which is not a model. The is analytics, right?
[00:38:34] Ken: Yeah. Oh, they, they do say, I forget what I read an article recently. They believe that their model is worth over a billion dollars a year to them, their recommendation system alone, which I think is fascinating.
Which I also find a little bit surprising, but, you know, the question is how much would a model that was that like you are, I could create your worth to them. Right. Is it something that is like, like w what is the difference between what they do and like the next step. Like, you know, w what, you know, from, from ground zero, maybe it's worth that much, but compared to another reasonable model, probably isn't that great difference, which I think is fascinating.
[00:39:17] Sheamus: Yeah, that's very true, but that's why I like recommendation models, because the end of the day, the keys in the name, it's a recommendation. I'm not, is that a binary model, right? Exactly. Like you'd be pretty pissed off of Netflix. If there, if they give you one movie to watch you touch you turn on Netflix.
We know you so well, here's our recommendation. So you're going to watch them, Emily and Paris, five episodes straight and nothing else. Which I wouldn't mind cause I gotta catch up my viewing, but do you know what I mean? So I'm talking, I'm talking about, about, cause think about financial markets or healthcare, right?
Financial markets. I mean, you're going to bet 50 million are not going to bet 50 million. I'm either going to allow this trade or not allowed to straight arm. You're going to tend to those patients that have cancer not have cancer. Right? So this is what I mean by as we progress into this as, as the stakes get higher, this stuff was going to become very, very important, very, very quickly.
And I don't think it's, I guess my main points and my prediction was it's not going to matter until it really means.
[00:40:20] Ken: I say, well, you know, it's funny, you should say that. So I'm reading this book right now. It's called the changing principles of dealing with the changing world order by Ray Daleo. Oh, very good.
[00:40:30] Sheamus: That's on my list. Tell me when you're done with it.
[00:40:33] Ken: Yeah, I will. I will. I'll bring it with me when I see an next I'll drop it.
There we go. That would be sweet. But, but the book, it talks about how, like cyclical, the financial markets have been over time and it's significantly more at a macro scale, predictable than a lot of us realize.
And so I think that that's something, you know, basically like big things happen around war. War is relatively predictable based on inequality within within country and other geopolitical forces and essentially every 75, a hundred years. I have these new world orders that are created based on essentially everything blowing up and going to scratch.
And then there are all these new opportunities and all these new revolutions or whatever that come about. And I think that maps really well to what you're describing is that essentially when everything hits the fan, that's effectively, when there's all these opportunities to reorganize and to create new industries or new roles or new positions or new companies.
And, you know, the challenge is generally pretty hard to predict those things that we haven't had a lot of history with, but it's also something that I think is, is, I don't know if it's necessarily a ground truth, but it's something that in my short time around on the planet, I've, I've seen a lot of truth too.
[00:42:07] Sheamus: No, no, no. That's an awesome point that I think your listeners should kind of take that to heart because one of my favorite sayings I say to myself, when I'm, when stuff just kinda collapses and falls to the ground is you know, when one door closes multiple other doors open, right? So, you know, I had a consultancy up until about 2018 and ODSC was going from an expensive hobby to something I had to run full time.
We've a amazing team there. You know Avara, Anna and tons of other people work in ODSC. So it's been amazing to work with them and. You know, we just got really bad timing. Like we just hit our stride in 2019 and then, you know St Patrick's day, thanks in Boston. They shut down all the bars and then like COVID started.
And all of a sudden, like overnight, it was just, it was just turned off. And I had, I've had three or four moments, not in my life. I was in London working for a hedge funds and the guy who was head of hedge fund called me up and says, James we're shutting the whole thing down because the financial markets collapsed.
Before that I was in college, I had a job and I was going across in the red line and then the.com bubble burst and the job I was supposed to take was gone. So, so it's just like, yeah, shit happens. It tends to happen to you. Cause that's what makes it global. Right? It affects a lot of people. So you got to learn to roll with the punches.
Number one, I really, really believe like when NAS, when, when, when one door closed this many other doors open, because remember it's opportunity cost. So it doesn't matter if you fail to get that university you you get fired from a job where you don't, you don't get the job you want, this are, you don't get the promotion.
You want it. It's like, it's really up to you because to me it's a mental state. So when that door closes, it's been closed on you. Right? Cause I'm sure a lot of your listeners are like me, but I like to beat the door down, but you know, you know, you bounce off a door so many times I get slammed in your face.
So to me that's always been a great lesson and I've really taken that to heart. So a great example is You know, I always try and surround myself with a great team of people, because I know when opportunity strikes, you don't know when it's gonna strike number one. And you know, when the tough gets going, when things get, when things get tough, you need a really good team of people who are passionate around this doesn't mean.
So, you know, doesn't matter if you've hired these people, yourself are their colleagues, you know, just it's really important. Whatever you do is just surround yourself with that kind of a team. And I'll give you a great example. So COVID shut us all down. But from April, 2020 until where we, you know, January, February 2022, we built our own virtual hosting platform for hosting our virtual conferences, you know, built our own from scratch, like a startup. Couldn't do that in a year. Addition to that, we built them an online virtual training platform, collab plus training. And it's now we built a career matching platform, KaliaI, plus careers.
So there was four things as well. I can't, I can't remember how many projects we did, but these are all machine learning projects. We built them with peer teams of people and it wasn't just me. It was this awesome team of people. Like I said, four of them all, Anna, Irene, era, I could name them all day, but everyone got behind these different projects.
We got them all running at once. They were super passionate about them and they've used an opportunity. So I'm like, don't be, don't be kind of depressed. That shit happens. It's common. It's kind of like, I like to invest in the stock market. I always tell my friends, I'm like, you got to have dry powder.
I'm waiting for the shit to hit the fan. So cause you know, then, then you've got opportunities, right? And that's when working in finance taught me is that you have to be patient, the opportunity will come and you just have to be ready for it. So it doesn't matter if that's the wife in life. You're looking for the perfect partner, I guess be ready.
I don't know. And in work in but stuff like that. So a lot of people are very, my optic, like, you know, this sucks why's it happened to me like Yeah. When the pandemic is just the person's story as well. I'm like I can sit there and be fat. I'm going to go run my eScience. Like I set myself a personal goal, which I failed to do, but I'm like, I'm going to lose 10 pounds while this pandemic hasn't been, I'm going to go outside with my mask on, go for a run every day.
So you can let events control. You are, you can take it to opportunities, events. And that's a really important life lesson took me a long time to learn that. But I do have a stubborn streak, but that's played well to my advantage. Like, and I'm also a super optimistic, you can't be a business person without being super optimistic, but being an, being an optimist is a pre-college question qualification to be an entrepreneur, but it can also be your biggest weakness, right?
Because you have to be optimistic. You have to believe in yourself and your team and what you're building. But then life will come along and smack you around the face. And then, you know, you've got to, at the same time, when you, when, when those punches lands, you gotta figure out, like you gotta have a plan to pick yourself up.
So I know I re I re I went on there a bit, but I think what Ray's talking about, it's like this stuff was coming and I think everyone can learn from the pandemic and what stuff went on prior. And I'm going to deal with that because, you know, I'm always thinking of myself, not negatively, but like what's coming next. You know.
[00:47:45] Ken: Yeah. Oh, I liked that so much. And you know, something that you said in there is really important to realize and To me, you described the patients, but you also described proactivity, right? So there's one thing to be patient and wait for things to happen. And there's one thing to be patient and slowly preparing for things to happen, which is the category that you fall into.
I think something that, that people like us, people that are optimistic, people that are, that view themselves as entrepreneurial are always trying to do is that they struggle to sit still. They want to be working towards something. They want to be building something or they want to be doing XYZ. And I don't think that that conflicts with patients, I think that it means that your, your proactivity is focused on preparing for something.
And then inevitably the patients is waiting for the right moment for, to actually happen and to get in and pick the opportunities.
[00:48:40] Sheamus: Yeah. Cause like you, I talked to a lot of people who are just starting out their careers and you know, my heart goes out to them because I can, I can tell that they don't quite understand the advice I'm giving them because like a lot of people who want to be in the data science field or want to get the next level, they're already, they're already working the job price.
I dunno. You could be flipping burgers. You could be working as a data analyst. You could be working as a marketing person on, you're trying to convince your boss, your company to put you into a data science role. Right. Are you're you're interviewing with it. It's on no, one's giving you a chance. So I say that I like pretty much everything I transitioned into.
You know, you're not, no one gave me a conference to run. You know, I had to go out and find good people, find a good team. No one told me I could work in quad finance. No one told me I could be a programmer. Like you have to work towards those yourself. And then you have to have the patience to give yourself time to build up the scale and have the perseverance to do that.
Because you know, I hear people on my, when I'm doing my own, you know, we do a kind of a career lab as well. And they're like, Sheamus, I've talked to 50 companies, I'm a data analyst and no one can be Brexit, data scientists. But I think there's an element of I have to do this and I have to do this now because I believe everyone's on a different timeline.
Some people can get there quicker than others, but you've got to keep chipping away at the break, you know, chipping away to break that down. And it will, if you have the patience and perseverance, it will come because sometimes it's about finding the right that the right strategy. Right. What's your end, stuff like that. So that's, that's kind of key to it. That's, that's a big piece of it.
[00:50:22] Ken: I really liked that. I think that all at least I found the opportunities come to you if you're starting to create them. And that gave me a really interesting way to view opportunity costs, right. So if I don't land a job and then I just sit on the.
The opportunity cost of me missing that job is virtually nothing, right? I mean, there, I'm giving up nothing by missing that job. On the other hand, if I don't land a job and I'm still pursuing different things and I'm working on my portfolio, maybe even try to start a company or do whatever the opportunity cost of taking that job right.
Is in theory, infinite, it's sorta like a Pascal's wager type of thing where it's like you know, I'm not a very religious person, but I think that, that the idea is interesting where it's like, if the rewards for believing in something are infinite, right? If there's infinite upside, even if there's an infant testimony, small chance that that the event happens, you should still.
Pursue that event because the expected value is infinite. Right. And I think there's a lot of, a lot of value associated with that in the job search. Is that okay? I didn't land this one position, but I have infinite opportunities out there if I still pursue, if I'm still moving forward, if I'm still building on what I have already.
And I don't know, I find that unbelievably powerful. So, you know, as we're coming to the end here, something I really want to touch on is how you built ODSC, you know, like where did, what's the origin story? How did, how did you build it to where it is now? Obviously I spoke at one of the more recent ones that I had a great time.
What did that progression look like for you?
[00:52:06] Sheamus: Yeah. So that's interesting. There's there's I know I noticed a simple answer, but I'm going to fail to give it that was, as you said, at the very start, a bit of an awakening and a lot of luck and a lot of collaborations. I think I mentioned earlier how I went this meetup, that was an epiphany for me.
Right. So I came from a background where in, you know, when you're doing consulting, when you're working in finance in place like that, you're really not allowed to talk to you talk about your work. And I really, really love this whole idea of like, I can go to a meet up this. Person's going to share their story.
Shared her learning shared with an audience. I didn't have to pay for us. I'm not being pitched anything and not been sold anything I'm there just to learn. So I love that concept. So I started the Boston data science meetup, and I was giving a few talks myself, you know, like yourself, giving a talk a lot of work, and I'm like, you don't want, it's much easier.
I'm gonna invite people to come talk. Right. So I'm started inviting people to come talk. And then I realize, yeah, this is awesome. I get to meet wonderful people like yourself. I get to learn about what they're working on. I get to meet other people that are interested with their various work about so kind of snowball from there.
And then I got together with a bunch of them. You know, meetup is very collaborative. Like a lot of people think meetup competes, we set up it's very collaborative. So we started to do the scene called the Boston data festival back in 2012 - 2013, before my five meetups together. We had stuff going on all over Boston in the evening.
You know, and different events space. And then as I said before, I've always known that when I, you know, when a billing working in a startup, advising a startup it's team team team, I know that people are sick of hearing that, but I'm like, I really want to do this. But the two is successfully.
I want to do it globally. I want to do it a scale. That will be of interest to people. So I was talking to a few friends and I met this guy Mamet Moody, who was running these early morning discos in New York. You know, those things in New York were big fat for awhile, you know, pre-COVID you go there to dance from them, you know, like, I dunno from eight o'clock in the morning, till 10, I forget the recalls, but he was running these things.
So I'm like, okay, I'm a data scientist. Matt is a great party organizer. I want this stuff to be fun. Cause I'm going to these stodgy conferences. I'm like, man, let's start a conference. He's like, all right, Janice. Let's do it. That's awesome. Yeah. He's like, how big is your party budget? No, but he was, he was my first kind of hiring that he was, he was unbelievable. He was fantastic. And then from there we just hired, we started hiring really good passionate people because you know, we, you know, to me, it's a dream job. We get to reach out to some of the most interesting, exciting people in the world of data science and AI, right.
To other people. Like if you're not in data science now you're like, Oh my God, this is so boring. Like why do I email this person and stuff like that. So we've had a ton of fun building a team of people who are really passionate about the whole concept of, of, of running a conference, you know, and to be part of that team, like you kind of have to.
It's you know, I built software and I love people using software. But as an entrepreneur, I had a restaurant business back in Ireland, like a decade ago as well. And what I used to love on a Friday afternoon and the place would be packed and be like 200 people in there. They're all having a good time.
People are getting drunk, they're getting happy. You know, they're with loved ones and it's kind of the same at a conference, right? You've got like five, you know, back in 2019. Now it's not like that you have five or 6,000 people all chatting, all connecting. And it's an unbelievable sensation. Excuse me. To know that you made that happen, that you brought all these people together and, you know, it's a really good kind of like shared team collected.
Like, you know, as a team, you know, you execute on this because it's a thousand points of minutia, right? So it's like it's like a premier league soccer game, rugby game, whatever your sport is American football game, where you spend all this time a team practicing, practicing, practicing, and you go out there and you deliver perfect game.
All right. You know, it's only a, see we used to have a few little fumbles in there, but.
[00:56:28] Ken: Well, you know, I think the thousand little points of minutia is really beautiful thing, is that what I've learned from the past couple of years, especially creating content is that people are what make any of these businesses successful, right?
Like even data science where we're coding and doing all this stuff at the end of the day, the goal is. Improve value and some humans life, right. at a microscopic level. And the more people that I meet, the more opportunities that I find are presenting themselves, whether it's to speak, whether it's business, whether it's just like making friends, whatever it might be.
I am a firm believer now that, that all of the value, all of this stuff is pointing towards building a community or improving the community or a part of. And I really like that overarching message. I mean, it's brought me significantly more happiness. It's also brought me some financial benefit over the last couple of years as well.
And I'm excited to see or to speak with someone who shares that sort of that idea that the collective is.
[00:57:32] Sheamus: And I will tell people who are like, you know, you're tired of your job. You're bored of your job and your secure job. You know, you've got to value your work time. You've got to value your home time.
But if you, if you follow your passion, maybe it's not even data science, but if you follow your passion, it rewards your multiple level because of, you know, success. As you were saying earlier, can, it's not just about monetary reward because you know, success is you get to a point in your career because you're passionate about it.
And people know you're passionate about it. You get to a point in your career where you can work with awesome people that are, that are as passionate to you about something. It just makes life just so much bloody, more fun and enjoyable. You know what I mean? And that's a key part of work-life balance because if your work-life is sucking the life out of you, that's affecting your home life, right.
That's affecting your personal life. And if you're working with people. That you love working with you love collaborating with you love creating what, it doesn't matter what field you're in. It's not worth. Yeah. Yeah, yeah. And you know, every my career, I didn't value that. I didn't value that enough.
And now I would, you know, I would take less money for a better team hand over fist because you know, life is too short and yeah. And if you follow that but what's interesting though. I think if you follow that road to successful, you know, that the work-life balance will workout. And you just have a richer life in general.
[00:59:03] Ken: Incredible. I think that's a perfect place to end. Sheamus, thank you so much. I enjoyed this as always a pleasure talking to you. One thing, how can people learn more about yourself, more about ODSC, any of those things? What's the best place to find out?
[00:59:18] Sheamus: Well, yeah, loads, loads, loads, loads, loads, loads there.
So want to reach out to me the best places LinkedIn, or you can email me, you can reach out to me on LinkedIn. If you want to attend ODSC and anytime ever just go on the website, we've got a free expo. There's also a career lab. So we have lots of speakers, much more qualified than myself talking about data science careers. Ken Jee, your fabulous host director spoke with that event. And San Francisco last year. So you can access that for free. So ODSC has got lots of free stuff going on. If you're expressing our careers platform, you can go do it. You can go to AI plus our careers. We have a free career lab there and all that good stuff as well.
So yeah, just, just sign up for the ODSC newsletter and you'll get information about all of this stuff we're doing. You'll have more enough. You can, after a couple of weeks, you'll be hit the unsubscribe button.
[01:00:12] Ken: I supposed to say that, but I'll link all of those things into the description below.
So, yeah, you'll have a link to Sheamus's LinkedIn and ODSC. Again, I had a great time speaking there. I'm so happy that you could come on the show and I can't wait until we talk again.
[01:00:28] Sheamus: Ken, as always, man. Great talking to you. Love your energy, love your show. So we shall chat again.