Eric Weber: [00:00:01] You have to have the right mindset if you're going to do big things, you kind of have to be willing to be tenacious and be willing to also fail. Eric Weber: [00:00:08] Like they're going to be uncomfortable because you're going to fail at some point. And that's not something people love to know, but it's something that's just the reality. Harpreet Sahota: [00:00:31] What's up, everyone? Welcome to another episode of the Artists of Data Science. Be sure to follow the show on Instagram @theartistsofdatascience and on Twitter @ArtistsOfData. I'll be sharing awesome tips and wisdom on data science as well as clips from the show. Join the Free Open My Slack channel by going to bitly.com/artistsofdatascience. I'll keep you updated on bi weekly open office hours. I'll be hosting for the community. I'm your host, Harpreet Sahota, let's ride this beat out into another awesome episode. Harpreet Sahota: [00:01:12] Our guest today is a lifelong learner who loves working with data, educating people about data science and helping them excel in technical roles through the development of his expertise in statistics, machine learning, Python and R, he's cultivated a passion for sharing his work and experience with others to help them become excited about data science, as well as educating executives on all aspects of data science. He holds a bachelors degree in mathematics from the University of Wisconsin, a master's in business analytics from the University of Minnesota, a master's in mathematics from Arizona State University, a graduate certificate in statistics from Arizona State University and a PhD and mathematics from Arizona State University. And if that's not impressive enough already, he's earned dozens, of course, certificates for MOOCs, such as Coursera and data camp. He's built a solid career in mathematics education, having held positions as a graduate, teaching and research instructor at Arizona State University, director of mathematics and science education at Oregon State University, and on the biostatistics faculty at the University of Minnesota. In the industry, he's held roles such as the Director of Data Science and Analytics for Health East, Senior Data Scientist at LinkedIn, principal of Data Management and data science at CoreLogic, and is currently the senior director and head of Data Science and Insights at ListReports. Harpreet Sahota: [00:02:27] In his current role, at least reports, he leads the business, intelligence, data science and research science teams and is involved in strategy, development, resource planning, as well as hands on technical development and deployment. His team's mission is to deliver easy to use World-Class Tools that seamlessly integrate into the daily lives of realtors. On top of all of his professional experience, he's been invited to speak on more podcast than I can count an innumerable number of speaking engagements and online seminars. Most recently, he's giving a keynote address titled Data Science is not about how many models he build at the 2019 Data Hack Summit, as well as a panel discussion on the topic of why 85 percent of A.I. projects fail. Alongside some titans of our industry, such as Terry Singh and Kunal Jain, when he's not crashing it as an acclaimed data scientist, he loves being outside, especially enjoying the tons of hiking and cycling trails all over California as well as travelling so that he maintains a healthy perspective of the world. Harpreet Sahota: [00:03:22] So please help me welcome me. Our guest today, a lifelong learner and rock star, our industry, Dr. Eric Weber. Eric, thank you so much for taking the time that you scheduled to be here today. I really, really appreciate it. Eric Weber: [00:03:35] Thanks. I'm thrilled to be here especially - And I think when I accepted the invitation, the world wasn't quite where it is now. I think I remember thinking, I wonder if I'll be free at 4:15 p.m. Pacific on that day. And now I can pretty much make meetings of anytime of any day because there's no commute. There's no it's a it's a strange time, but it's also a really good time to be able to talk about our industry. Things happen in people are feeling a lot of pain and anxiety and stress. And I think it really helps to have perspective on what it all means for us as a community. And so I'm thrilled to be here. Thanks for the invite. Harpreet Sahota: [00:04:17] Oh, yeah, definitely. And happy to have you. And that's an interesting way, how all of a sudden the world just got so much smaller because we're all in this weird thing together. And, you know, having the technology to keep us connected in this way is pretty awesome as well. But I'd like to to talk about your journey into data science. Can you talk to us about how you first heard of data science? What drew you to the field and some of the challenges you faced breaking into the field? Eric Weber: [00:04:43] Yeah, that's a really - it's a good question and one I think that is often potentially overlooked because even with data science, people talk about it as if it's pretty new. Like they're like "oh data science is still..." It's actually been around for a number of years, but people's journeys into it continue to amaze me. They're all different. There's no one set way to end up in a data science position. For me, I was in the academic world teaching statistics and programming, experimental design, things like that, all the way up until 2013, 2014. And I distinctly remember in a phone conversation with my dad and he at the time was an engineer for United Health. And he was like, well, they're talking about all of this big data, stuff like blah blah blah, like, don't you do things with data? Like I do. But I don't know if I do things with big data. And that kind of kicked off my fascination. I think I remember after that that was the first time I drove into some of the online certificates. And when I started doing some of the MOOCs that you see on my profile with like Johns Hopkins, it's like, what is this all about? And I learned in a lot of ways that the concepts and the ideas that I already knew and used were pretty foundational to working with data. For me, it was having to think about how to operate with data at scale, whereas I was used to being able to run things on my local machine. Eric Weber: [00:06:20] Right. I can run things in R. I was programming in R when there was nothing pretty about. It is a brutal experience. Now I look at it and it's see this beautiful tidy, clean thing to do. Back then it was not. So for me, my journey was all about figuring out two things. One, how to work with data at scale. And two, what does it mean to actually do data science in a business context? And those two things are really, really important. And they also, I think the probably overlooked my transition really entailed a lot of learning about what it meant to really do data science in a business context. You can learn about data science and how it operates within business. But to actually make it effective for business is actually quite a journey to go from an academic side where you're in a classroom and you're used to delivering things in a way, and then you push them over the line and they're done almost like a homework assignment in our industry. It's sort of continuous and that continuous learning, that continuous improvement and the fact that even when you create projects, you never really get to just turn them in as homework. That was sort of an eye opening experience for me. Aside from the day to day business to go away from taking my teaching mindset from the classroom to my team, that was a huge experience for me. Harpreet Sahota: [00:07:55] Are you an aspiring data scientist struggling to break into the field within Check-Out dsdj.co/artists to reserve your spot for a free informational webinar on how you can break into the field? This going be filled with amazing tips that are specifically designed to help you land your first job. Check it out at dsdj.co/artists Harpreet Sahota: [00:08:21] Yeah, it's interesting you say that pushed over the line and submit like it's homework. When I was a biostatistician, I felt like I was doing homework all day long. I just like submitting assignments to the government, it was like... Yeah, it's totally different on the other side in business, like you said, everything's continuous. But I'm curious, where do you see the field headed in the next two to five years? Eric Weber: [00:08:42] I think I see the field head in in a non uniform direction. And I think that is maybe the most valuable thing that we have going on for us is data science has evolved into what it always probably was meant to be is a bunch of sub-disciplines, just like the idea of saying that you're a data engineer or that you're an engineer overall. There's so many different types of engineering. They all require different skill sets. There's very few people who are experts in everything. So as much as there's been a lot of debate, I think people like, do we need specialists or generalists? [00:09:16] I think we're getting to the point where specialties are the norm. But that's not a bad thing. Eric Weber: [00:09:21] Just because you're focused on time-series work, generally speaking, doesn't mean you're not skilled. Just because you tend to focus on machine learning or system design doesn't mean you're not skilled. It's just infeasible at this at this point to hire someone who's good at everything. That isn't just from a candidate perspective, though, it's from a company perspective. They're figuring out that what they've done in the past, which was hire data scientists and basically make them responsible for all things data. It's actually really tricky to figure out how to use them effectively. This is produced in a lot of companies reorganizations, merging parts of data science with engineering, parts of data science with product. In some cases, data science has become its own entity within an organization. And so for me, what's coming next is this alignment, because companies are at the point now where they're like, how do we get our ally out of this practice? And a lot of cases getting that out is making sure you figure out that there actually are sub-disciplines and you can't hire one person to do the job of three, even if they're extraordinarily talented. Eric Weber: [00:10:29] And so there's a simultaneous change like students and people prepping to go into the field are changing. Companies are changing in how they hire and what their expectations are. And I think in a lot of cases, it's actually that this really interesting experiment to see what's going to happen with people that are working with data at some point, it still is like this sexiest job, but company is still have a bottom line and more so than ever with the current public health and economic conditions that we're in. Companies are going to be evaluating the true value of essentially everything within their midst. And while in the last 10 years it's perhaps been easier and they've had budgets that allow them a little bit of wiggle room for the next 12 to 24 months. That's not going to exist. And so I see it as like a real prove it time for data science. As much as it's hard to think about that. It really I really think it's going to be that in the next year or two. Harpreet Sahota: [00:11:32] So kind of that vision of the future. Well, what do you think is going to separate the great data scientists from merely good ones? Eric Weber: [00:11:40] Flexibility like this idea of being flexible doesn't mean that you can handle a whole bunch of tasks coming at once. It means that you can sort of ramp up your - not just flexibility. I think the best way to put this is, you know, what's required to do different tasks. And you don't always use a uniform approach to do everything. A data science task, task A is probably always going to be different from data science task B. And they're probably going to require different skills. They're going to require different models. They're going to require people to understand how much is actually needed to solve the problem. You don't need to build an incredibly powerful model for every situation, but you need to know what's going to allow the business to thrive in a productive way. The other part is businesses, like I said, are going to be looking for the bottom line value of data science as a practice. And I'm saying this specifically, if you're not talking about research groups and teams that are at Google or Amazon and all of these places that will continue to be research focused, they may not be the immediate business value. And almost every other case data scientists to understand that their skillset gives. Eric Weber: [00:12:54] Value to the business by actually delivering and business value and not just scientific value. It's going to be incredibly important. This I think there's a disconnect. Sometimes people go into data science positions thinking that it is doing a series of projects and to establish really good code. And the optimal solution. But then they kind of hand it off as if the business is going to magically use the thing that they've created that I don't think is a good assumption. This flexibility to pick the right approach or to also know why an approach may or may not work and your ability to really transform the business with your solutions. Those two things are going to differentiate data scientists who are going to stick around at companies and data, scientists who are going to being viewed as really scientists. There's this weird perception. Oh, they just do the science-y stuff. If a company is looking at you as you do the science stuff, I can probably guarantee that they don't see how you're impacting the business. And those are really key things for people to keep in mind. Harpreet Sahota: [00:14:01] What I got from flexibility is like the ability to operate comfortably with a compass instead of a step by step roadmap. I'd love your LinkedIn headline. I learn every day. That's freaking awesome. Can you talk to me about the importance of being a lifelong learner and how you've adopted that to be such a big part of your identity. Eric Weber: [00:14:25] I think it goes back, and some people ask me a lot like, why did you go to school so much? And the degrees are one thing. But I think if you look at school and any program as a chance to, like, form habits, it's a much better illustration of its long term value to you and your career. This idea like coming to class or coming to work, whichever it may be, ready to go and ready to do things that you haven't done before. That mentality, I think, kind of is what drives me to think about my work and to think about the value that I can provide. I think we often go to school and we think about getting a degree that we continue to improve enough and do enough until we get granted an education, right. Or we get grants and or certificate. But in almost every case, those skills you learn are going to be outdated. And in our field every two years, probably, probably less. And it's much more about bringing the energy and bringing the desire to go into situations where you don't know how to do things perfectly and being uncomfortable. And so if I really put it, I want like my goal is probably to be uncomfortable most days with something like that. Mentality is a good illustration of the scientists component of data science. It's not just a series of skills that you deploy on well-formed problems. Eric Weber: [00:16:00] You should probably be working on stuff that isn't clear. It isn't very easy. And that to me is why data science often tends to be more highly compensated position or deserves to be in some cases, because you're willing to push outside the comfort zone and solve things that people don't really have a good answer for. Harpreet Sahota: [00:16:17] That's really what all the growth occurs, right? Isn't that discomfort when you're really stretching yourself? People understand that physically when you work out your muscles. Yeah. Like your muscles are going to grow when you are stressing them, but we can't see it inside the brain. Happening right now, they don't make that correlation. What's your advice to aspiring data scientists who feel like they have not learned enough to start applying for jobs? Eric Weber: [00:16:44] I think that's a - one, that's a fantastic question. Two, I think there's never going to, no one's ever gonna go ever going to know enough to solve every problem. Eric Weber: [00:16:56] It's more about the things I mentioned before. Do you have a diverse and deep enough skill set to tackle problems with different approaches? And how do you know that? How do you assess what solutions a problem needs? And also, like, are you willing and hungry to deliver for a business and not just write a script and call it good enough? Because in my experience, writing a script and producing a report tends to have actually pretty, very little business value. It's all the actions that come after that, as much as people want to imagine that that report is going to get picked up. Eric Weber: [00:17:34] And implemented. It's not going to not unless you're the one pushing. And so it's less about do you know everything or do you know this mountain of information? It's more about do you have a variety of skill sets? Do you understand how to work on different problems with different approaches? And are you ready to demonstrate that you can solve a problem for that business? This is why I think for most interviews, a huge amount of the value is that you can show that you understand how their business operates and how you potentially have a lot of value for that business. It's less about checking the boxes on. Can you build a gradient boosted tree? Can you train a neural net? Do you understand what back propagation is? That stuff is potentially useful, but to me it's about do you know when to use it and why it might be helpful? And does the business problem you're trying to solve actually meet it? If it doesn't, then be willing to say so, because otherwise everything you do is going to end up being trained on a neural net. And that's just not going to deliver the value you want for the business. Harpreet Sahota: [00:18:44] So kind of the opposite question in a sense. What's your advice for those data scientist who think that they've learned enough and don't need to learn any more to be successful? Eric Weber: [00:18:56] Well there's two possibilities. One is that you've learned enough so that you can do the daily tasks that are asked of you in your current position. Eric Weber: [00:19:07] But have you learned enough to help the business change? And actually ask questions in reverse? Right. You can you might reach a plateau on a business where you're not asked to do anything harder than a certain level. Eric Weber: [00:19:20] And my question on that point is, OK, are you ready to leave? If they're not going to continue to push you or are you ready to contribute and learn enough so that you can push them to another level to do things that they couldn't really imagine doing before? And so there may be enough like you may. And again, this is a I think this is a pretty personal decision. You can probably determine that, you know, enough to get by. But getting by is not a long term solution to delivering value for a business, because what you're doing right now to get by is probably going to be automated in a few years, or there's going to be a tool that makes it super easy to do what you do. And so then it's a question of, OK, what are you doing that the rest of the business that that somebody else can't. And most often that is having formed relationships and knowledge about the business that other people could not do, even if they were the most technically skilled person on the face of the planet. And that's where you start to see businesses willing to retain people because they know that if I lose that person, it's going to be a net negative on them. And that typically is where people define themselves as their ability to transform things on the business side. Harpreet Sahota: [00:20:47] What tips do you have for a data scientist in a team environment who's maybe scared of looking like they don't know something but does want to openly communicate that to their teammates? Eric Weber: [00:20:59] One, you got to get over it like there's I really don't like giving. I try to be gentle on some cases, but in this case it is a if you don't know something, there's one or two ways it comes out to one. You ask somebody and maybe you risk in that moment not feeling like you know everything. Or two, you do something without asking people. And because you didn't ask. You miss out on something super important or you don't do something up to the expectation level that they have. And then the feedback on that case is super negative because you've turned, you generate a solution that is not sufficient. At the same time, you've also illustrated that you don't know what you're doing or you miss something. In either case, you're going to have to in some way demonstrate that you don't know everything. It's probably easier to do that to ask somebody else on your team. Eric Weber: [00:21:55] But this also comes down to interpersonal dynamics. Dynamics are hard. There are teams where asking questions may just not meet the norm. And there are teams where asking questions is well accepted and supported. Eric Weber: [00:22:09] I tend to be happiest and I recommend that people try to find teams where asking questions and being open is is valued. That doesn't mean that your shoulder tapping on somebody every fifteen minutes because at that point no one's gonna be able to do their job. But it does mean that part of what you assess for the value of a data science position and whether you're a good fit is are people you working with - Are they open? Are they willing to chat? Are they willing to teach you things here and there? And that's, and if you're uncomfortable doing that and scared about it, that's also the time where you should really be talking. This is where a good leader should be setting the tone. And where I see a lot of teams fail. Is their dynamic is often a reflection of their leader or lack thereof. And so as much as it's about, are you like in my opinion, you just go for it, ask questions, be open about it. If that's not the norm, then maybe long term, we're not a good fit for that team. But you learn that either way. Either way, asking questions is critical. Harpreet Sahota: [00:23:15] Yeah, definitely. I feel like it. It really is your responsibility to just be vulnerable. Right. Even if it goes against the norm of the team, it's on you to just be vulnerable and own up to it, unafraid to talk about your shortcomings. Eric Weber: [00:23:29] It's uncomfortable, too. It's definitely uncomfortable, but you get better at it as time goes on. You're gonna look stupid. At some point in your career, I'd look stupid many times in my career. And that's just it's just the norm. You're going to eventually look silly or ask questions that don't make sense. You're not always gonna be the expert in the. And if you are, you're probably in the wrong room. Right? That's probably not a room you want to be in at all. Harpreet Sahota: [00:23:59] Yeah. So do you consider data science machine learning to be an art or purely a hard science? Eric Weber: [00:24:07] I think science in general is an art. So I think it'd be really hard to - I think any science done right requires technical mastery. But also there's a whole lot of gray area in how you do things and the choices that you make. This is why in science there are people who are good and there are people who are great and there are people who are probably legendary because even though they might all have similar technical skill sets, it's often about how they ask questions about how they answer them and how they are pretty relentless. Eric Weber: [00:24:38] That kind of defines people. So science is taken the way that I think about it is definitely an art. And in so many ways, like data science is an art and something that is cool is that as you become better at it, you start to see that there's a lot of ways to approach problems. There's so much open space that's not clear, even though you know all of the algorithms. It's not always obvious what it what makes data scientists A better than data scientists B. It's not in the model that they build, typically. It's in like how they actually define a question and then pursue that answer. Harpreet Sahota: [00:25:20] What role does being creative and curious play in being successful as a data scientist? And how can someone who doesn't really see themselves as creative be creative? Eric Weber: [00:25:31] Yeah, that's a good question. I mean, with data science, I mean creativity. I think creativity is often assumed to be this sort of wishy washy thing, like how do you define what creativity is? And it's not about your ability to construct some beautiful canvas from nothing. Creativity is probably actually about a lot of small decisions. And like when you're tuning a hyper parameter on your tuning something else, when you choose what models you use or you choose how to, like, create a particular feature in your dataset. Others may not. Right. Creativity there is often. It doesn't look like the type of creativity that we tend to think about in general. Creativity is also how you posed the question, are you very often people are really talented or asking different questions than people who are just starting out. So creativity...People are like "I'm not creative". Well, I'm not creative in like 90 percent of my life. I go to the same store, I go to the same restaurants. I don't like change it up that much. But when it comes to, like, how you practice and how you show up and play in data science, like creativity is about the small decisions, not like some grand art piece that you're trying to design. Harpreet Sahota: [00:26:52] I kind of defined it as just putting old things together in new ways to solve come up with like a novel solution to something. Right. So we talked a little bit about how, you know, in our previous roles with kind of homework-y. Right. Pushing something over the line in submitting it. But what challenges do you see data scientists who are always looking for a hard and fast rule, or step by step recipe to follow possibly face when they're working outside of a homework or study environment and they're in the real world where things are way more ambiguous. Eric Weber: [00:27:32] And I think the challenge is that step by step doesn't work super well in the real world. Even like the scientific method seems really well-defined and like, OK, we do this and we do this and we do this. But really, it's like a sequence and back and forth and a sequence of testing and hypothesis development and testing and hypothesis development. And it's iterative. So I think part of what people are not used to is like there's a continuous cycle of development when it comes to doing things and data science that is not present in homework, typically like when you set a random seed and then you're deciding what you're testing and training data is like. That's step one. And then you are doing subsequent steps from there all the way to where you present like the quality of your model. But in practice, like data science has so many other components to it. The solution you develop may not work for the business. The solution you develop may have issues because the data that you sampled from was not representative of all of the customer base. There are a whole lot of potential problems. And I think part of being in data science is that when you're doing the homework version of data science, you're not necessarily used to... A lot of the normal issues you face are abstracted away from you. Eric Weber: [00:28:59] You get to focus on the stuff that fits. Step one, two, three, four and five. But when in reality, it's like all the stuff that falls outside of that is the real problem when it comes to doing most of this field. So you learn like - so much of our homework and our education focuses on like the 10 percent of the stuff that you're going to like. You spend 90 percent of education on 10 percent of the stuff you're going to do in the real world. And I think that's a hard adjustment for people to make. They're like, oh, wait. But I learned how to do all this stuff. I learned how to use. I use learn how to use NumPy. I learned how to use keras or tensorflow. I'm like, I'm like, that's great. If you got to that point, that means you've done something awesome. That probably means that you spent a lot of hours on it. I think what we talk about it and it's really popular to say 80 percent of your time is spent working on getting data in the right format. I don't think it resonates with people until they're actually sitting there doing it. Oh, my gosh. This is actually terrible. Harpreet Sahota: [00:29:56] Yeah. Yeah. Like going from broad data. Yes. Jumbled up Rubik's Cube and then having to form it up into, you know, a nice cube with nice, you know, colored sides and everything. Changing it up a little bit now here. Harpreet Sahota: [00:30:10] So what's it mean for you to be a good leader in data science? And how can an individual contributor embody the characteristics of a good leader without necessarily having the title? Eric Weber: [00:30:23] I often differentiate between like a leader and a manager because they're not always the same thing, a leader. And so even if you are in charge of people, it doesn't mean that you're a leader. Eric Weber: [00:30:34] It means that you perhaps are giving people tasks. I look at leadership. Here's an interesting book. It's called Multipliers. And it's a really, really good book. I think it is my my personal. I think a lot of it resonates with me because being a good leader is about figuring out how to unlock, amplify, and develop the people around you. Eric Weber: [00:31:02] And in most cases, that means that you're not very often you're not going to be the most technically competent person in the world, you're not going to be an expert in every area. But you have to figure out how to create that environment where people can actually develop and want to want to not only wants to be a part of routine, but like it's mutual. You give them something and you help them develop, accelerate, get so good at something that they're ready to move on. And at the same time, they give you extraordinary work. I think in that book and a lot of other cases like, you know, a good leader in organizations, when you ask people like a team matching phase. Right, like a Google or Facebook or any of these places, if you ask them what team should you work on, you should work on this person's team. And that's typically a good indicator that that person does something over and above. Just managing it typically means that they're actually accelerating their career of the people around them. And so when I think about data science leaders, whether it's a manager or someone who is has a team of five or five hundred. It's all about are you accelerating people around here from an individual contributor perspective? So many. There are a lot of leaders who are individual contributors. An individual contributor, I think has the stigma that it's like you only do things for yourself when in fact most of the things that you do that are impactful are about changing the people around you and changing the systems around you. Eric Weber: [00:32:40] Right. If you look at most most companies, how they evaluate leadership is your impact and your impact as an individual contributor can be huge. You can make it easier for your whole org to do something. You can help develop the skill set of those people around you. You learn how to do something in a better way. So the title to me, well, titles are important in some way because they signal things about level and what your responsibilities are. Like your title isn't always going to tell you if someone's a leader or not. You really have to know more about working with them Harpreet Sahota: [00:33:15] That's a great book, Multipliers. I've had the misfortune of working for a demand issue. You know, immediately after I worked for my diminisher, I worked for an accidental diminished. When I took on this other role, I had the opportunity to just become the multiplier myself and try to embody those characteristics. And in my current role, I'm definitely working for for a multiplier right now, and it's so refreshing to have somebody who you just want to give all you can to to help bring the initiatives to fruition. Eric Weber: [00:33:51] I think it's hard for people to understand is if you're in the right environment. The multiplier leader creates an environment that people want to be a part of. It becomes less about your compensation and less about your title because you actually are just enjoying what you're doing. It's hard and technical skills because we're taught that technological skills are kind of the defining feature of a data scientist. But when it comes to organizational impact and team growth and team cohesiveness, someone has to drive the boat. Someone has to...And that doesn't mean that they have to push everybody along. But it means that they have to like, generally orient people in the right way and let them do what they're good at. And I think that is a hard thing for people to understand. Harpreet Sahota: [00:34:36] Speaking of the technical skills, a lot of up and coming data scientists tend to focus primarily on the hard technical skills, and they think that that's what's going to separate them from the rest of the crowd. What are some soft skills that candidates are missing that are really going to separate them from their competition? Eric Weber: [00:34:52] I think this can be a this is a tough one, because when it comes to interviewing, people gravitate towards the hard skills because they're cleaner. Right? They're like, am I good at SQL? Am I good at applied stats? Am I, Can I do leetcode like these things are more measurable in some way. Soft Eric Weber: [00:35:14] Skills, though, are much more about my way there. They do a couple of things. One, they apply to connect with people. Right. And this doesn't mean that you have to be a people person. This just means that you have to have an approach ready so that you can demonstrate that you're someone who people would like to work around. That genuinely means sometimes just asking people about their interests. What do they care about what motivates them? Like, if you turn it and say, what would you want someone to ask you? What would you want to share about your work? Like it's OK to ask people to share those things and to really talk about those things deeply. So while it's a soft skill, it's like forming connections with other people. Even if you're not a people person, it's still really important because it makes them feel like they can imagine working with you as a teammate, your ability. And I think a lot of people like, well, public speaking skills. I don't necessarily think it's public speaking skills. I think it's the ability to be clear, your clarity of communication. There's a lot of cases where people say a lot of words, but they don't say anything. It's like so it's uncomfortable but record yourself when you answer a question. I think a good example. I think there were a couple of major MBA schools this year that in lieu of the in-person interview, they had candidates record like a one minute or two minute response to two questions that they didn't know ahead of time. Eric Weber: [00:36:46] I think it's a really good mentality. Are you able to communicate something useful in a clear way? Can you do it within a two minute window? Cause that's what's gonna be asked of you in a business context. So can you communicate about something in a relatively concise manner that's key. And we all like to think that we know everything. But then when it comes time to actually explaining it, some people just go on and on and on. And that's not what's needed in a business context. So getting to the point, making people feel connected with you. And part of getting to the point is also making sure that you can communicate about the ideas, complex ideas, in ways that make sense to people who may not be well versed in your field. Right. Using a lot of words to a lot of people talking about machine learning, you can literally say any word you wanted and say machine learning and they would get the same message out of it. You have to make it mean something to them. And those three things, I think are really key when it comes to what we what we would call soft skills. Harpreet Sahota: [00:37:57] It's funny you were saying how people use a lot of words, but don't say much of anything. I just tell my wife that because every morning the prime minister of Canada, Justin Trudeau, I think 11:00 a.m. comes out and gives a national address to Canadians regarding the COVID situation. And every time I'm like, dude, he just is using a whole bunch of words to not say anything. Eric Weber: [00:38:18] Yep. We I think a lot of cases politicians have thrived on the ability to say very little while saying lots of words. And it's but it's tough. You see, at times like this where people are stressed and they're in crisis and they're anxious. The ability to say something meaningful in a concise way is a pretty sought after skill set. And it's not out there, but like a lot of people just are not good at it. Harpreet Sahota: [00:38:48] And so for those data scientists out there who haven't yet broken into the field, how could they develop their business acumen and product sense? Eric Weber: [00:38:59] There are a lot of there's a few ways. I mean, there are a lot of companies that post and host their own blogs about the problems that they work on. And very often the questions they ask will be focused on those same problems. So if you think through some of the things that these companies are tackling and think about how you would approach them, think about why you would approach them in that way. Eric Weber: [00:39:23] There's a lot of problems that are transportable between companies when it comes to like for SaaS companies, like customer churn. Stuff like that is going to be sort of a universal problem that they face for products and metric tracking. That's going to be relatively universal. Events may be different, but the concept of product analytics is data science is pretty consistent. So you can prepare for a lot of product and marketing and sales data science by just doing it for one company, because the questions that you ask there are probably broadly applicable for more specialized roles. Eric Weber: [00:39:59] That's where understanding what the company is doing exactly and how they're making money is really important. I recommend that you think that you look at the presentations of people that people from that company have given their all over YouTube. They're all over the place. The power points are sitting out there. There's video all over the place. Look at some of the case studies that they've been working on. At least gives you a sense for the type of data and what they're dealing with. You can't prepare for everything, but if you have thought through a lot of different cases, I think it's really going to be really beneficial to you. Harpreet Sahota: [00:40:32] It's really good advice, actually, taxing the advice that I give to the mentees that I'm responsible for. It's pretty much the exact same thing you said. Harpreet Sahota: [00:40:46] What's up, artists? Be sure to join the free open mastermind Slack community by going to bitly.com/artistsofdatascience. It's a great environment for us to talk all things data science, to learn together, to grow together. And I'll also keep you updated on the open biweekly office hours. I'll be hosting for our community. Check out the show on Instagram @theartistsofdatascience. Follow us on Twitter @ArtistOfData. Look forward to seeing you all there. Harpreet Sahota: [00:41:15] What advice or insight. Can you share with people breaking into the field who are looking at some of these job postings, some that seemingly want the abilities of an entire team wrapped up into one person, and they end up just feeling dejected and discouraged from applying or even trying to enter the field. Eric Weber: [00:41:32] You just have to shine with what you're good at. And you're not the one who's going to make the determination about if you have the requisite skill set for that job. They're going to be the ones who determine that. So getting dejected is basically making a judgment before someone's actually paying a judgment. There are a lot of jobs that people apply for that they don't have all of the skills listed because a lot of cases, those companies figure out pretty quickly that no one like that actually exists in the real world. They just don't. It's companies often will post things because they've seen their competitors use a similar job posting or they've just aggregated all the words from all the job postings that they've seen. It doesn't mean that they know what they're looking for or that they're going to actually like minimum requirements for five years. Doesn't mean that they're going to hold to that. It's that you have to have a PhD. You know like a lot of cases, those things are put up there to prevent them from getting 10000 applications. Eric Weber: [00:42:31] But I think you have to be comfortable with what value you can deliver. If you think that your skill set and your background can provide value for that company in that position, then go for it. There's it's not a frustrating I mean, it's a frustrating thing because job search is imperfect and the interview process is imperfect. Eric Weber: [00:42:56] I feel the way more interviews than I have succeeded in making. That's true of almost everybody at this point. You've interviewed enough and like even the best people end up failing in or not getting called back. And that's just the reality. It sucks. Job search is not a pleasant thing. It's much more about your tenacity. And if you let yourself get down with it, it's going to it'll have a net negative effect. Like you gonna feel yourself getting down and take a pause until you're not. Otherwise, you're not going to be able bring your energy on a daily basis. Harpreet Sahota: [00:43:28] What are some challenges that a quote unquote, notebook data scientist is going to face when it comes time to productionize a model? And do you have any tips for them to overcome those hurdles? Eric Weber: [00:43:41] Most of the time when we train and build models, we don't think about their scalability and we don't think about their scalability within within a production level system of competing for resources with other things. Eric Weber: [00:43:55] We don't necessarily think about their latency. We don't think about how long would it take to serve up a recommendation to an individual user. You kind of take the input data as given, when in fact the input data may actually be streaming in real time. And so there's some. And so for me, it's not it's about timing. Mike, how does this model work within this within the ecosystem of that product and also scalability? Do you have the resources to be able to actually operate that model in a reasonable time frame? A user is not going to wait 10 seconds for a recommendation to be made. It might you might be willing to wait 10 seconds, but they're not going to. And so production is in many cases about not just creating the model, but building it, deploying it. And how do you maintain it? How do you know that something's going wrong with it before you have thousands of timeouts happen because it's taking too long and your users are pissed? You have to really think carefully about. And this is like people are like, well, what is a system designed for data scientists? And I think we don't we don't have to think about it. Generally speaking, for decently sized companies, because the engineering teams tend to worry about the actual production and platform work. But at the same time, you have to build things that have a chance fraction making it there. There are so many things we take for granted feature engineering and data ingestion, training and deployment. We control a lot of things locally that you don't control in a production environment. And understanding what those things are and how to handle them is really important. Harpreet Sahota: [00:45:37] Oftentimes when you're developing a machine learning model, the intervention, the therapeutic you're building is your model itself. And then you've also got to assess the impact that therapeutic once it's out there doing its thing as well Eric Weber: [00:45:51] Mostly like most of the time We're trying to we're trying to either influence or understand behavior if you are trying to influence behavior. If you then influence an influence in a way that makes your model unable to continue to positively affect things like you, you think about it, you train your model on a certain subset of data, right. Ranges of input, variables, stuff like that. Eric Weber: [00:46:12] Let's say that the model you create changes people's behavior so that the input data coming in or the recommendations that need to be served don't fall within that range of the model and how you built it originally. It's almost like the model can create a scenario where the model doesn't work well anymore. And that's weird and not intuitive, but it's not like you get to see the model every time it comes through and gets trained again. You have to actually build the model to sort of retune it goes into production and you also have to be thinking about what the effect it has on your user base. What are the guardrail metrics that you care about for your users that are going to signal if something's going well or not going well? Because if you're doing something at scale, something bad happens for a lot of users. That's a big problem. You can't just like can't just unroll that very easily. Harpreet Sahota: [00:47:09] So last question before we jump into a lightning round. What's the one thing you want people to learn from your story? Eric Weber: [00:47:16] I think it's say you have to be pretty tenacious. You have to be tenacious about like if you want to be good at something. If you want to be skilled and you want to make an impact, he kind of we're going to have to do things that make you uncomfortable. I started posting on LinkedIn two years ago somewhere on that ranch. I was terrified of posting on social media like I iJet, generally speaking, don't like social media. But I started sharing things and people find a way to resonate with it. And it becomes super powerful and amazing. If you just think about it in the right way, but you have to have the right mindset. If you're going to do big things, you kind of have to be willing to be tenacious and be willing to also fail. Like, you're gonna be uncomfortable because you're going to fail at some point. And that's not something people love to know, but it's something that's just the reality. Harpreet Sahota: [00:48:04] So jumping into quick lightning round here, what's a topic, academic or otherwise outside of data science that you think every data scientist should spend some time researching on. Eric Weber: [00:48:15] Social science, man. Eric Weber: [00:48:17] I think understanding the world in which you live and operate and the dynamics and behavior of human beings, that's at the end of the day, who are really dealing with the most cases. If you don't understand the broader context in which they live and operate and the stressors present there. It's hard to understand who your users are and what you're building things for. It's huge. It's important. Something that I wish more people did. Harpreet Sahota: [00:48:38] What's your favorite question to ask during an interview? Eric Weber: [00:48:43] It depends on the interview. But I generally really like to know about a time when someone not just failed, but they had to admit failure to somebody else. And I don't care if it's in a work context or not. It doesn't matter to me like. And you will not will not believe the number of just fascinating responses that get us. Because people this attaches very emotionally to people and they're often very willing to share this. But it tells you a lot about how they respond to something negative that happens, how they share something that is, by its nature, uncomfortable, but how they not only shared it, but a lot of cases they learned from that. That's one of my favorite questions. Harpreet Sahota: [00:49:32] What's the weirdest question you've been asked in an interview? Eric Weber: [00:49:39] I have... Eric Weber: [00:49:40] There is a... I'm trying to think about exactly how it was phrased, but it was something about the improving the situation there. How would you build an algorithm to judge hotness of something? And I'm like are we building on Facebook back in 2004, like, this is really weird, guys like. And it was with this really small company. And really, like I could tell at that point, it was like not gonna be a professional experience. I wanted to pursue that. They're like, what features would you use? Like, this is super strange. Let's not let's not go there. Eric Weber: [00:50:12] And but it tells you a lot. The questions people ask tell you a lot about who they are. And that one told me enough that I was like, all right, I'm good. Harpreet Sahota: [00:50:20] Interesting. What's the number one book you'd recommend our audience to read and your most impactful takeaway from it? Eric Weber: [00:50:29] I think Blink is probably the one that has resonated with me for a long time, mostly because it talks about how we make decisions and how those decisions are often not logical. And the way in which we process information in order to do something is really fascinating. Eric Weber: [00:50:45] The amount of effort we give to certain things and lack of effort we get to others and how our brain either makes complex decisions, simple or simple decisions, complex. It teaches you a lot about like human cognition. And it always reminds me that, like, even if we make assumptions about what our users are going to do, we're usually wrong. We just don't know. We just don't. And that's really hard. Harpreet Sahota: [00:51:12] Have you read that? Thinking Fast and Slow by Daniel Kahneman. I think Malcolm Gladwell in Blink builds heavily on on some of the arguments that make in that book. Eric Weber: [00:51:20] Yeah, yeah, yeah, yeah. Harpreet Sahota: [00:51:22] Both are really good books. So I don't know if this is still true for you, but I've you know, in a previous interview of yours, I heard that you, like me, are very early riser. Eric Weber: [00:51:32] Yeah. Harpreet Sahota: [00:51:33] So it's still four a.m. Eric Weber: [00:51:35] It's usually it's more like five now. I sleep in now. Weird, it's weird. And like you got a tech companies and they're like 10 a.m. breakfast. When I get up at five guys, like, I don't know what you're doing. Harpreet Sahota: [00:51:46] What's what's your morning routine like. Eric Weber: [00:51:50] I get up and I usually like I forced myself before I allow myself to have coffee and breakfast. I just move - 20 minutes of doing something right. PushUps, SitUps. Doing so, especially now where you're not moving that much on a daily basis. Like it gets your brain ready to go. I apprehensively opened Twitter and the news to see what especially these days, I feel like I'm opening this like horror show of stuff from like what happened in the last twelve hours. But I like to get my mind active. I like to know what's going on in the world. And I really like to also just take some time to be like, what is my day going to look like? What are the big things I have to hammer out? And if you keep those things defined every day, even if you don't finish them, I think you make some serious progress against them. Harpreet Sahota: [00:52:42] Have you read The 5AM Club by Robin Sharma? Eric Weber: [00:52:45] Yes. Yeah. Yeah. Harpreet Sahota: [00:52:47] Do you do you follow the 20/20/20 Formula? Eric Weber: [00:52:49] No, not really. I, I more or less just I kind of make sure that I get up at 5:00 and by 7:00 I have to have done some productive things and I basically just start to my day that way. Eric Weber: [00:52:59] Sometimes I have noticed that I'll get off track, especially recently I'll just be reading about everything going on in the world and like what what it was like. What have you been doing for the last last hours? Harpreet Sahota: [00:53:11] It's it's crazy how much time slips away. Harpreet Sahota: [00:53:13] You don't even notice it when... Eric Weber: [00:53:16] [Inaudible] It's really weird. Like, how have I been sitting here for two hours. Harpreet Sahota: [00:53:22] What motivates you? Eric Weber: [00:53:24] You know, that's people ask me that question a lot. And I just I like to know more. I like to know more than I did yesterday or whatever, that sometimes that's in work, sometimes it's not. Sometimes that's knowing more about myself, which is an uncomfortable thing to learn, and it's just like I'm doing something or I know or understand something I didn't quite get yesterday. And I think sometimes that small, sometimes that's big. But I tend to like to think about that. If I had a history of defining that, that would be great. But I think when I go to bed in my car, I'd learn this today. That is kind of what led to my LinkedIn tagline in the first place. Harpreet Sahota: [00:54:05] I love it. So if we could somehow get a magical telephone that allowed us to contact 20 year old Eric, what would you tell him? Eric Weber: [00:54:13] Invest in Amazon Eric Weber: [00:54:18] It would be. It would actually probably be like do what you're doing, because I knew I was headed for the academic world at that point. But if I hadn't gone to the academic world, I probably wouldn't have ended up in this one. I probably would never have thought about it. Like, I originally wanted to be a high school teacher, math teacher. And I still I love teaching, but if I hadn't gone to academia, I probably would be in a secondary classroom right now. And it was my journey through academia that made me go this direction. And so I just do what you're doing. I mean, there's probably some exceptions to that. I but I made some bad decisions when I was 20 that I would probably also be like, don't do this when you have the opportunity. But generally speaking, I'm lucky that I took the path I did. Harpreet Sahota: [00:55:02] What's the best advice you ever received? Eric Weber: [00:55:04] Pretty simple. That, like, be humble. Like if you're humble about what you do, it naturally brings people to connect with you. It naturally. And it's not just about connecting with people. It's about keeping yourself humble. As soon as you start thinking you're the best. You're in a position that is not going to be good for you long term. It is just not a good thing. And so stay humble, not just about yourself to other people, but like actually be humble in your own mind. Realize that you're never gonna be the most skilled the way you do. If you think you are not hanging out with the right people, Harpreet Sahota: [00:55:35] How can people connect with you? Harpreet Sahota: [00:55:37] Where can they find you? Eric Weber: [00:55:38] LinkedIn i the best place. It's easy. If I have a personalized invite message, it's much easier because there's a lot of invites that come in where it's just like, I don't know the person, but, you know, tell me who you are. Eric Weber: [00:55:49] A little bit about you, what you want to learn by connecting and go from there. [00:55:54] Awesome. Well, Eric, thank you so much for your time. I really appreciate you taking time on your schedule to be here on the show with me. Yes. This was a wonderful conversation. I enjoyed it. Eric Weber: [00:56:03] Thank you. I appreciate it a lot to stay safe and stay healthy.