Or applying to positions because that’s when you’re getting the feedback from the market, that’s going to tell you, what skills are or gaps you have on your resume that are keeping you from getting a job in industry.
This is HumAIn. A weekly podcast focused on bridging the gap between humans and machines in this age of acceleration. My name is David Yakobovitch, and on this podcast I interview experts in sociology, psychology, artificial intelligence, researchers on consumer facing products and consumer-facing companies to help audiences better understand AI and its many capabilities. If you like the show, remember to subscribe and leave a review.
Are you curious about the future of work and work 2.0? Will you even have a job in the future from a co-working space or remote? How about what you should learn to stay relevant in the market? On this episode of HumAIn I speak withKristen Kehrer and we get fired up about college scandals, professional development and if data science is even required for AI. We’ll discuss about the data swamp life you live with devices and messages. And when data goes down, what do you do next? Tune in now.
Welcome back everyone to the HumAIn podcast. This is David Yakobovitch, your host, and today our guest is Kristen Kehrer¹. She was a LinkedIn top voice in 2018. Has been a data scientist for over nine years in the industry, working with a host of companies, including: Constant Contact and Vistaprint. And today is very involved in education, keynotes and democratizing data science for all.
Kristen, thanks for being with us.
Oh my God, Thank you so much for having me, David. I love your podcast voice.
Thank you so much.
Am I allowed to say that?
This is so kind of you. One thing of being in education , since we’re both there, we learned so much about getting involved with audiences and training and teaching. One of the big things I’ve discovered, so for our audience, Kristen and I worked together on course development and supporting students who are doing programs with Columbia University.
So we’ve been able to see a lot and particularly it’s been in remote learning environments. And one of the start off with a story, for everyone today, about the future of learning and the future of work. We’re moving into a world of work 2.0, where no longer are working offline in an office. But now it could be a home office or coworking space, like we work, or these new digital home offices that are hybrids of our home and work.
I used to work in New York and Florida, and even smaller coworking spaces, but I don’t anymore. Yourself as well as I do, we both travel a lot, and I wanted to hear your thought of leadership between the office and home. Do you think the coworking business model is dead?
I can only speak from my own experience that when I first went back to get my master’s degree in statistics, I started that in 2007, I had gone back into academia. One, because like the housing bubble was bursting and I was gonna lose my job because I was working in the real estate industry. But I went to grad school saying to myself: I never want to commute to work again. And then I finished my master’s degree and found myself commuting to work.
My next job after that was still commuting to an office. Then I was still commuting to an office and it took me a couple years to get to that place where I could work fully remote. Even though you often hear of people saying that data science is a great place to be if you want to work remote, because there are opportunities, but it’s not. There’s opportunities for the people who are established in the industry and have already built a reputation. Once you’ve built a reputation and you have some experience, you can sort of name a lot of your conditions for how you work and how you go about it. But I do worry when we promote people who are trying to get into the field that it’s this beautiful, there’s unicorns and sunshine and everyone works remotely. Cause I don’t necessarily think that’s the case, but I do absolutely love working remotely.
I find that I’m more productive. I feel that I’m able to manage my schedule better where I am traveling a lot to speak, and I’m taking podcasts in the middle of the day, such as this one. So I don’t know if I answered your question. I do think that the industry, that education, and that working is going to move towards being remote and I am seeing more opportunities. But that it is the people who are more well-established now that are able to grab those. That it will open up more to others in the future, but that we’re not there yet.
We’re starting to see companies across the world, who are having these hybrid cultures of work remote for three days a week, come to campus. A few years ago, this was a notorious counter-culture, where Marissa Meyer at Yahoo said: everyone’s recentralizing, coming back to the campus, being in-person for this collaboration. But after that acquisition by Oath, a lot of that culture has started to reverse itself and been more weight. Maybe our mental model was not in that right direction.
So, we’re seeing companies that both succeed and not succeed in that regards. But I like what you’re saying, Kristen, that this translates not only to the future of work, but also to the future of education, and what those modalities of learning are. We’re both very involved in the space, both with in-person and online modalities for learning. And one question I’ve been curious about is the modality. If you’re in person, if you’re online, is this a question about access? Is this a question about learning preferences? What do you think there?
I just think it’s becoming life. I have two kids. I want the same opportunities that other people have. I want to be able to attain a work-life balance with some sense of being able to, get my groceries, keep my house clean. And even if I’m not like out shopping for groceries during the day, but I have them delivered from whole foods during the day. Just like I now do all my shopping online. The costs of attending a university, the biggest expense is staying in a dormitory.
So why am I going to, or maybe that’s not true of every university, but it is certainly a huge percentage of the cost to, actually, live there. Now, are there benefits to living there; would I prefer to be in an office if I had an office in my basement and I could see people every day? But like when I have to weigh that against an hour commute to work each way, which is two hours out of my day, and our lives are just getting busier and I want to take on more projects. I have more hobbies. I want to be there for my kids. It just makes sense to work remotely.
And like you were saying with Yahoo, like I have seen over the years, the articles saying: people are moving towards being remote. And then you see different companies saying like: okay, that’s not the case anymore, we’re bringing it back. And then a couple of years later, you’ll see articles saying: Nope, we’re going back to being remote. It’s a model that just works for people and the lives that they’ve built.
And thinking of the lives that we’ve built, and the interesting thing is about education. As you described, I went to undergrad at University of Florida, very well renowned for its sports. It’s academia, it’s extracurricular activities, 1000 plus clubs. And my attraction to the school was a combination of both. Yes, I want to get involved with these extracurriculars like tennis, and chess, and ultimate frisbee, and all this diversity; but is that what students should focus on? Should that just be the vocation? As I’m here to learn data science, I’m here to learn software engineering. The reason that I bring this all up is I agree with you, actually, the modality of learning is changing to our lives and the lives are continuing to digitally transform, but there’s a lot of issues going on with colleges now.
I’m sure you’ve been aware that this year the biggest college scandal to rock colleges, since doping in sports, years ago, has occurred. I am actually shocked. As a first-generation student for me to see about the privilege that people have chosen to get their students into colleges. It’s going to rock the educational world, and potentially even propel bootcamps and online learning even more in the right direction, because people are going to say: I don’t want to be a part of this.
I’m also a first-generation student, I would never have wanted to give up that experience that I had. And there was a lot that you get in terms of, even if it was specifically related to. When I was in my undergrad I was a math tutor at the math and business center. Maybe that isn’t something that I would have had access to if I was learning remote. And for my children, I certainly hope that they go to a campus somewhere, because that it’s a worthwhile life experience, that really, I did a lot of growing up in college.
Particularly, where I see online learning being the most useful is for professional development. The person who’s working nine to five and wants to have the opportunity to take a Columbia course, like the one that we’re helping to facilitate. Those options should be available to people because the world is connected now.
So there’s no reason why we shouldn’t have back, but I don’t want to discount the experience of being able to learn on campus and build those real relationships with the professors. Of course, I have students that come to my office hours and they ask me: how do I start a machine learning project on my own? How do I find one for my portfolio? Or I give them career tips, and some people will come and work to build those relationships. But I don’t think it’s as readily accessible as when you’re walking by the professor’s office and it happens to be office hours time. So you pop in and you start to ask those types of questions.
So, there absolutely are cons. There’s both pros and cons. But where the pros really lie is not in that 18 to 22 demographic, where the majority of people, that’s what their time is going to be there. That’s what they’re devoting themselves to being a student, furthering their education. But once you get into that, 24 plus, where you’re probably starting to think about supporting yourself and you’d like to further your education. In addition to that, that’s where online learning becomes just the best options.
With this scholar’s scandals that are coming out. The questions that, you and I are both thinking, you have kids today. I’m planning to have kids. And should I be saving a quarter of a million dollars in the bank for each one of my kids to get into USC, to get into Harvard? If it’s all worthless and other privileged individuals are going to get in.
It’s a conversation. It is a dichotomy of these two personas that you just mentioned, the 18 to 22, and then the 24 plus. Is 18 to 22 is affirmative action necessary, or do we need to bring it back? And where is that lining up with accountability at the university level. But I get also every day by students: why don’t I attend a program at Galvanize, Flat Iron, Medis, General Assembly, insert any bootcamp name here for 20 to $30,000, and just skip college altogether.
I don’t always have the right answer for them, because I agree with you. I had this very impactful college experience that has defined who I’ve become and sounds like it’s defined who’ve used become as well.
It meant a lot to me to be, just like yourself, first-generation. I was the first one in my family to graduate from college. That was a big deal. And I enjoyed having that experience. What you were saying in terms of saving for education. When I talked to my financial advisors, when we first had kids, they were like: you want to be saving about $700 a month per child. And that’s if you wanted to be able to pay in full for a private university in 18 years. Which was my goal. I went in there saying, like: I want to be able to pay for a full-time university.
And so then it becomes. Yeah, I don’t know, I get so nervous about what I should be doing with my finances, because I don’t know what the landscape’s gonna look like in 18 years. And with these scandals, I really don’t know what the road is going to be. So then it becomes, so should I be investing in my children’s college education, but maybe not in a fund that can only be used for education. Maybe I should be putting that money someplace where it can be leveraged differently if the world happens to change. I don´t know
18 years is ,of course, 18 years. The one thing that is definitely going to happen is there’s going to be more accountability and there’s going to be more availability of the information in professional development and knowing what you’re getting yourself into, and those criteria.
Whether we’re looking at the 18 to 22 year old demographic or the 24 plus, for both of them, it’s not just in the learning can there be accountability, but in what they learn can there be accountability. So that means a lot of the training that you and I are both involved in, which is very focused on data science and artificial intelligence.
And as a sideway of this topic we’re bringing to light is how the industry is evolving for adapting to implementing code. Now in 2019, some of the code projects are no longer: Let me build an algorithm from scratch, but let me use what Google or Facebook or any of the other big things already made to solve a problem. And I wanted to hear from you, what do you think is both good or bad about systems like AutoML?
My warcry has always been that machine learning isn’t a black box. These models are interpretable. You may not understand the math that’s underneath. But we understand the math that’s underneath and we can interpret these models or we can at least talk to them. And when we use these autoML type solutions it actually does become a black box. Because the company that is giving you this, or you’re buying the software from them, they’re not going to tell you what the parameters are, whatever, because that’s their secret sauce.
There’s a couple of different things going on here. So, for certain models, If you just need to understand who’s at a high likelihood of retention or high likelihood of attrition, then maybe a black box is fine. I don’t know that you necessarily need someone to manually build you a logistic regression model, if that’s something that you’re able to get with a couple pushes of a button. But if you want to understand what factors are driving that retention issue, then, maybe, you do need to build it manually. And I completely forgot the third. There’s like a third piece to this that I was thinking about. Even as we do go to automate a lot of this stuff, one thing that will never be automated is those people who are able to come up with creative solutions to business problems that are not common across the industry.
So there’s always going to be problems that are unique to that business or something that they want to solve. And that’s where the data scientists of the future are going to really differentiate themselves. The people who can say: Okay, I’m not just building another logistic regression model, or I’m not just going to use random forest here. But the people who are able to say like: Okay, I’m going to use an off label use of this algorithm in an interesting way and solve a problem for the business.
There is a great actual book that I’ve been reading, I recommend it to anyone who’s on their journey in data science, AI, or in understanding what’s going on?, without getting too technical. So there’s this book that came out recently, it’s called Interpretable Machine Learning by Christophe Molnar.
And it’s regularly updated, a very interesting book because it’s about making black box models explainable. And one of the big things Christophe talks about is exactly what you just mentioned, Kristen. Is how common is the problem. How has it already been solved and how transferable is that across an industry. The challenge is as products that we’re using each and every day start implementing AutoML or auto AI solutions.
We almost lose our humanity in that everything’s just being automated. And where does that leave, actually, the role of data scientist and role of analyst and reporter? Who’s going to monitor that?
So a lot of the systems that they’re coming out with, have capabilities in place that will tell you when your model is no longer performing at a certain accuracy, and maybe you need to go in, and maybe that there’s been a structural change in the inputs that are going into your model. One that just comes off the top of my head is in 2010, 2011, when the price of oil really spiked, and people started transferring like crazy to natural gas, the models went all wonky.
And that’s something where you look at it and you say like: Okay, there’s some factors that we didn’t think about. And we’re absolutely going to need analysts who have business context and understand how to play with data to, sort of, rectify those things and maybe find new economic variables that are now. Somebody still needs to decide what the best inputs are. I do worry about people putting inputs into a black box model, cause we’ve all seen the output of stepwise regression and you see that sometimes it comes back and it suggests a variable being there that just makes absolutely no sense.
So there is a need for people who have this contact and understand how variables interact and the underlying mathematics. Just not sure to what extent as these solutions continue to become better. I’m not sure to what extent, we’re going to have to have that context. I literally don’t know, David.
What I’m hearing from you is that these solutions are going to become more automated and there will be complex tasks that humans will still need to be there for. But this begs the question.
I have a really close friend, who’s been wanting to get into tech for awhile. Let me share with you the story. He’s been wanting to get into tech, traditionally studied finance and has been involved in that industry for many years, but has not figured out the right way to launch the path himself into a career in tech. This individual may be, actually, quite similar story to many listening to the podcast today. He’s thought about going to community college.
Thought about doing a master’s in data science program, considered a bootcamp part-time online, full-time in-person. He’s considered this, but he actually asked me a question this past week. He said: Is data science required for AI? Should I just be focusing on AI, at the first point? And that question put me in an interesting place, because it’s exactly what you just said, Kristen. We’re not exactly sure where the industry’s going.
Data science, the beginning of machine learning is sort of the foundation. I feel that helps you get into those more advanced algorithms. And maybe that’s just because that’s the way that I brought up. Maybe somebody could go into AI without having the foundational stuff. I don’t want to speak for everybody in generalize. But at the same time I also wonder too. We see more applications of AI coming out.
But at the same time, at this point, the majority of companies still have their data in a data warehouse. A lot of the use cases are the. So we have this software that’s allowing for AutoML. I don’t know that the adoption is there yet. So a lot of the stuff that does need to be built right now is really manual.
So for your friend, I’d love to know if he is at the point where he’s tried applying to get in somewhere? Cause I do see a lot of people that take a lot of courses and they keep learning, and keep learning and the learning it’s never ending. It’s completely endless. And at what point do you Stop to learning and start applying to positions. Cause that’s when you’re getting the feedback from the market. That’s going to tell you, sort of, what skills are or gaps you have on your resume that are keeping you from getting a job in industry. Because I’m sure there’s some people out there who could just learn a little sequel, have a bachelor’s degree, get a job as an analyst and ,sort of, maneuver and move and shake their way into the position that they want to be in.
So should you go and become an expert in deep learning before you ever apply to your first industry job? I don’t know. That seems like a hard sell to me. If I was looking for somebody who was highly technical, but didn’t have any industry knowledge. Like that might be the way to go, might be through what they said in the old days: getting in at the bottom and working your way up.
I couldn’t agree more on that. And when I talk to students daily, the question is: am I looking to get an entry-level job? Yes. Okay, fine. Let’s scope that out. What do you mean by an entry-level job? I want to be a data scientist. Okay, where’s your skill level today? Excel and a little bit of SQL.
Great, perhaps your entry-level job should be data analysts, market analysts, financial analysts, business analysts. Exactly echoing what you just said, Kristen.Then maneuver and working that way up unless you have a lot of time to study. The challenge in the industry is you go to a master’s in data science program without having that industry experience. And then, why off the get go would Amazon, Facebook or Google hire you? They want to see that experience. So, the reason I’m sharing this, is today we are living in what I like to call the data swamped life.
We’re living in a data swamped life between devices, between information, between messages that are so distracting that does not allow us to pause and get any calm in our life. I remember, in one of my first analyst jobs, when they used to work for a big bank, I would always get pinged by communication from my colleagues. And before you know it, I would look up at the clock and it would be 5:00 PM at the end of the day, and I had not accomplished a single thing I had set out to do for the day. Whether it was learning a new technology or implementing a solution for the team, I would get distracted.
And I’d love to hear from your thought leadership and industry any recommendations you have for new data scientists or those working their nine to five jobs and have it be more effective and efficient to accomplish what they’re setting out to do.
So, before I get into that, I want to go back to what you said before about people finishing a master’s degree and wanting to become a data scientist. And I just want to say Amen.
I’m so with you. I felt like a couple of years ago when, or several years ago, when the term data scientist was coined, I felt like it had this air of this is not an entry-level position. And somewhere along the line, as the years have gone, people have started to think that they can graduate out of school and get right into a data science position. But then at the same time, we don’t have a clear definition of a data scientist. And if you ask 10 different people, you’d get 10 different answers.
And some people would think that the person who is working in Excel, using SQL and maybe built one model in R is a data scientist. And that there’s a ton to do, to sort of, put a stronger definition around what that is. But my feeling over the years is that the term has taken on a different meaning. And then of course, there’s been this flood of people trying to get into the industry, and as that’s happening, data’s getting bigger. The tools are getting easier. Things are becoming more automated. It’s a really exciting time to be in the industry and then…
You’re so right, Kristen. Is such an exciting time to be in the industry. A new report comes out every day, calling data the new oil. And because there’s so much data out there and tying what you just talked about these newly minted data scientists who get their masters in data science or finish a bootcamp program.
And thinking back to that case that I just mentioned. When I was an analyst and getting so distracted by being pinged by communication and not knowing how to be effective in the role. I wanted to know, What have you found to be successful to help you accomplish what you’re setting out to do? In a day by day work for someone new.
So, there’s basically a million different things to this. First of all, I would like to say that today I had some work that I knew I needed to deliver today. I blocked off hours on my calendar, so that I could not be disturbed. No one could hop on my calendar at that time.
I also evaluate the meetings that I get on my calendar. If it’s not something that I need to be at, I remove myself from that meeting. But in addition, I also want to say that I’m pretty good at inviting myself to meetings that I should be in, but the business doesn’t necessarily think that they need me.
They don’t know that they need me, but they need me. If you’re designing a hypothesis test, and it’s a group of marketers that are coming up with this test, like: Please call me, and I’d like to give my input. There’s also a lot to do in terms of workflow, and these pings, maybe it’s that: Hey, file a ticket and we’ll worry about that during prioritization and I’ll get back to you, and let you know how this ask is prioritized against the other things that we have in the pipeline. So that you don’… I’ve used the analogy before that sometimes you feel like a short order cook, except it’s more like, would you like K squared tests with that. Instead of, would you like fries with that? And that’s a position you can get into and it involves multiple things.
First of all, the directive needs to come down from the top that the data science is going to focus on the highest priority or the objectives that are going to end result in what they think will be the highest ROI. And those should be the things that we focus on.
The ask that comes from the business that does not align to those. We need to have people who are higher up, especially for the newer data scientists, that say that coach them through this: Hey, by the way, if you get an ask from the marketing team that doesn’t align to this priority, kindly tell them to file a ticket that’ll go in the backlog.
So it’s a whole mix of, because people say they’re data-driven. But at the same time, we do have a lot of these analytics and data science teams that are more reactive than proactive because it’s missing in the strategy. They don’t have a true data strategy. They don’t have the true support from executive leadership that is setting things up in a way, so that you don’t get accosted by the constant asks for: can you build me a model to do this? Can you pull this data for me? I can’t get at it.
Being data driven starts with being people driven, and it starts from the top down. We are all here about being a data-driven culture. A lot of the work you’re doing is about bridging that gap, through keynotes and podcasts and teaching.
But I wanted to share another story that might be interesting about how we’ve become so attached as a data-driven culture. What happens to us as a society when data goes down? Just this year, there was a rumor that Facebook had a denial of service day. For many of us consumers here in the podcast today, you may have been afflicted where you can no longer use your Facebook, your Messenger, your Instagram, your WhatsApp.
For me, I was one of those affected, and I did not realize how dependent I was on this technology to communicate with other people, so much so, that it was a significant disruption to my business activities and my personal activities during that day.
I remember when I used to use AOL and MSN and MySpace, all these different apps over the years. Today we’re seeing how our lives are very much controlled by Facebook, WhatsApp and Instagram and even Snapchat. The very fascinating most important point when all these services went down this year is a very little unknown service outside of the tech space got 3 million new users that day.
And that service is called Telegram. For those of you who haven’t checked it out, Telegram is a WhatsApp type communication that is all about privacy and all about decentralizing their servers. And I found that so fascinating that in this data-driven culture are there alternatives, and perhaps they’re starting to emerge in the market.
I shared us all with you, Kristen, both out of our shared mutual experience, but then also wanted to hear from you about trends and signals that you’re seeing in the industry. In this data driven industry we’re living in that are moving from the fringe to the mainstream that you’ve identified. You might want to share with our audience.
For me, the biggest thing about being a data obsessed culture is that now in industry we want every single decision to be made on data. And I don’t think that that’s necessarily the right outlook.
If you have a strategy and you are going to make a decision, regardless of what the data says, like that is not something that should be taking up analyst or data science resources, that’s a waste of time. If you already know what it is that you’re going to do. So that kind of ties into what you were talking about is that.
Everything that we do requires data, is capturing data. We’re trying to use data for everything. I don’t think that data needs to be used for everything, if it’s not actually going to inform what it is that we’re going to do.
Informing what we’re going to do is just the start of a conversation like this, Kristen. As we continue the work together, perhaps you and I will have more aha moments with the students and the organizations we collaborate with, and for all our listeners on here. This is what HumAIn about.
It’s about bridging the gap of humans and machines by thinking as a data first culture where humans can be at the forefront of tasks of projects and about making our life one where we are working together. Kristen, it’s such a pleasure for you to be with us here on the show today.
Thank you so much for having me, David. This was fantastic.
That’s it, for this episode of HumAIn, I’m David Yakobovitch. If you enjoyed the show, don’t forget to click subscribe on Apple podcasts or wherever you are listening to this. Thanks so much for listening and I’ll talk to you in the next one.
Data Moves Me LLC