Kirill: This is episode number 377 with Chief Data Scientist at Metis, Dr. Deborah Berebichez.
Kirill: Welcome to the SuperDataScience podcast. My name is Kirill Eremenko, Data Science Coach and Lifestyle Entrepreneur. Each week we bring you inspiring people and ideas to help you build your successful career in data science. Thanks for being here today and now let’s make the complex simple.
Kirill: Welcome back to the SuperDataScience podcast everybody. Super excited to have you back here on the show. We got a very cool episode coming up. Today’s guest is Dr. Deborah Berebichez, who is the chief data scientist at a company called Metis. If you haven’t heard of Metis before, this is a great company that provides all sorts of boot camps and trainings to individuals and companies in the space of data science. Deborah is the chief data scientist there.
Kirill: She structures all those curriculums and helps enterprises understand how to best train their teams. What we’re going to be talking about today is going to be separated into two parts, the first part is actually going to be quite different from data science training, it’s actually going to be focused on women in STEM, a very important topic of how women can get into and succeed in the fields of science, technology, engineering and mathematics. Deborah is very passionate about the space and I think everybody should be at least aware of the space or even a better case scenario, do something to help make this space as welcoming and as amazing for absolutely everybody who wants to enter this space.
Kirill: Deborah’s going to share her story and her passion and she has actually lived the story coming from Mexico and always wanting to study physics and science and having bias against her because she was a woman in these fields. I found it to be a great story to experience the emotions that somebody actually goes through on their way. If you want to educate yourself in the space of women in STEM, definitely this first part of the podcast is going to be very helpful in that space. You’ll learn about mentoring women in STEM, and what Deborah is doing in that space. The fixed mindset versus the growth mindset, being brave, not being perfect and if you’re a business owner how you can actually promote a safe and inclusive working environment in the spaces of STEM such as data science, in your organization.
Kirill: In the second part of the podcast we’re going to be talking about data science and data science training. We’ll touch on topics such as critical thinking, true data literacy, training curriculums for companies, the future of data science, and data science silos, especially that last part about data science silos and what that means not silos within an organization but what kind of silos will emerge in the data science industry in the future in the coming five or 10 years. I found that very insightful, a very vivid description of the future of data science that we can anticipate. Very excited about this podcast, lots of valuable information, both from a community perspective and also from educating yourself on the topics of data science, and I can’t wait for you to check it out. Without further ado, I bring to you, chief data scientist at Metis Dr. Deborah Berebichez.
Kirill: Welcome back to the SuperDataScience podcast everybody. Super pumped to have you back on the show. Today we have special guests calling in from New York City, Deborah Berebichez. Deborah how are you going today?
Deborah: Hi Kirill, thank you for having me on your podcast.
Kirill: Super excited to have you. First of all, I want to ask you, Berebichez you mentioned the surname was Lithuanian. Tell us the story behind that because it doesn’t sound Mexican at all.
Deborah: Yes, my grandparents were from Lithuania. They were Jewish. They escaped when the conditions in Eastern Europe were not very good especially for Jewish people and so they actually tried to go into New York. But New York City I believe closed the immigration doors in 1929 or so because so many people were coming from Europe, the boats came to Ellis Island in New York. There’s a museum there and then they were put in quarantine. They gear down eventually to go to Mexico which was at the time this Golden City, especially Mexico City was just this beautiful city that needed hardworking people so they opened the doors and people came there. They probably thought, “We’ll make it to New York in a few years. We’ll work hard here, and then we’ll move up north.” But the conditions were so good, they created a community and schools and just the second generation was able to go to university already and Mexico was really noble and it treated people very well at the time so they formed that.
Deborah: My last name was originally something like Berebichic, something like that. But when they immigrated to Mexico in the registration port in Tampico they switched the name, the ending to E-Z. Trying to make it more Mexican like Paris, Gonzalez [crosstalk 00:06:04].
Kirill: Sanchez.
Deborah: Yes, exactly. But of course, it didn’t really work well and so now it’s this mishmash of Lithuania and Mexico. Nobody knows exactly how it should be pronounced.
Kirill: Also have you done the DNA test for family history?
Deborah: It’s a fascinating thing, and I should do it. I know my dad did it before he passed away and I have to check. He did it with 23 and me I have to check if I’m able to access his records. Because he’s, of course a family member, but because he’s deceased. May he rest in peace. I don’t know if I’m able to legally access because I would like to know where we’re from and collect all the historical information for my family.
Kirill: Yes, for sure. You should do it as well. It’s very interesting. You’d have a very diverse pivot tree. Mexico, at what age did you move to New York?
Deborah: Well, New York is maybe 12 years ago but I moved to the US way before that. I grew up in, as I shared part of the Jewish community in Mexico City, which was a very nice place to come of age, but it was also a very conservative place both Mexico and the Jewish community tend to be quite conservative. I was told from a very young age that I shouldn’t consider a career in physics and math which were things that I was curious about. I was told that since I was a girl, I better pick something more feminine like communications or art. I felt like I was rejected by society and I had to change my desires to know about the universe and the world. I had to always hide my dreams of becoming a physicist. It was not until college. In fact, after two years of college in Mexico, that’s when behind everyone’s back, I applied to schools in the US so that I could become a transfer student and continue and pursue finally my dream of becoming a physicist.
Kirill: What got you so excited about physics? Why were you into it so much?
Deborah: I don’t know. I was an extremely inquisitive child. I still am. I ask questions about everything. I’m very curious about why the world works the way it does. I’ll give you an example. My husband is also a physicist. He’s a professor and so we spend endless hours discussing why that effect happened with the water on the table or when our kids do something, we try to explain it through what I call physics glasses. We just put on those principles and critical thinking skills that allow you to see the world in a way that wants to be generalized so that you can say which is exactly what physics does. It abstracts certain principles that are applicable everywhere in similar situations. We’re able to say these happen because of the angular momentum, or our son slipped in the kitchen because of the density of the particular oil that my daughter put, whatever.
Kirill: Interesting. With this interest in physics, you were at the same time back in Mexico told that this is not a career path that you should pursue. How did that feel?
Deborah: I learned Kirill with time to hide my love for math and physics. That made me very insecure. I thought I had absolutely no skills. I remember that I did take calculus in high school, but it was not only at home that I was told my mother told me, don’t ever let the boys know that you like physics because there’ll be intimidated and you may not be able to get married, it almost happened. But I was told also by the teachers in school, and then the counselors that were hired to tell us what career path to take after high school, all those told me you’re very curious, you should study philosophy because it asks a lot of questions. What I had learned is that my desire to do physics was not something that I could just put down and hide somewhere. The hunger to know about the world in mathematical sense was not going to go away.
Deborah: I would read historical books about obscure physicists like Tycho Brahe, a Danish astronomer who apparently lost his nose in a duel because he was not a very sociable person. But nevertheless, he did incredible observations of the sky and from these observations that Kepler apparently stole at some point, the laws of planetary motion were derived. He did incredibly useful data science and physics research that gave us one of the most incredible insights about our universe. I thought maybe I’ll be like Tycho Brahe. Nobody will like me. I’ll be locked up in some observatory, but I’ll have my science and my observations with me. I had my silent heroes like that and then when it came time to go to college, I went to philosophy in Mexico.
Deborah: I was sure that was going to be it that I wasn’t going to be able to move because I knew that universities in the US cost eight times what they were costing in Mexico City at the time. I was going to a private university and my parents probably couldn’t afford sending me to a university in the US. After two years of philosophy behind everyone’s back, I started applying to schools in the US because I had learned that in the US you could do more than one major that is a liberal arts education, which would have allowed me to also venture into physics. What I did is I applied to the schools and I got an email back from Brandeis University, a small grade school in Massachusetts that said you have great essays and very good grades and we can feel your passion coming out of your essays. If you take these extra exams and write mission statement about why you want to do physics you’ll be considered for a scholarship.
Deborah: They actually ended up giving me a full scholarship to attend Brandeis University. Within a few months my life completely changed. I flew to Boston, I have never seen the snow. I flew in the middle of the winter and I arrived at Brandeis and it was the beginning to a completely new life. I’ll tell you the story because I think it’s a good story for your listeners. I decided to have the courage to take a very generic course my first semester in astronomy because I was fascinated about the night sky and the stars and how the planets move and all that. I was very intimidated by physics and math because I had been told for many years that I couldn’t do it. I sat in the very back of the class.
Deborah: There was a teaching assistant who was a graduate student in physics by the name of Rupesh. He came from India and Rupesh and I became very good friends. We would walk around campus and discuss well beyond the topics of the class. I would ask lots of questions. I would want to know everything about thermodynamics and quantum mechanics and planetary motion and so on. Rupesh told me, you’re not the typical student that wants to just get good grades and do well in the homeworks and whatnot, you really have a desire and a passion. That’s the most important thing. You should consider maybe going into physics and I said, no, but I only have two more years left in my scholarship because I was a transfer student and I can’t pay to stay an extra year.
Deborah: One day, we were walking in Harvard Square and I had tears in my eyes. We sat under a tree. I looked at Rupesh and I said, “Rupesh, I just don’t want to die without trying to do physics. I’ll probably not succeed. But I want to give it a try.” Rupesh got up and called from a payphone. We didn’t have cell phones at the time, he called the chair of the physics department who was his advisor, Dr. Wardle and he said, “I have a student here who only has two more years at Brandeis. She just got here as a transfer student. She would love to switch from philosophy to physics or add it as a major.” The chair of the department called me into his office and he said someone else has done this many years ago Ed Witten who by the way is the father of string theory. Very famous man clearly a genius switched at Brandeis from history to physics. We know that it can be done.
Deborah: I was like, “He’s pulling my leg comparing me to Ed Witten.” But he handed me a book, which at the time was an alien language for me. The book was called Div Grad and Curl Vector Calculus in Three Dimensions. He said, “If in two months by the end of the summer, you are able to master this material, we’ll make you take a test and we’ll let you skip through the first two years of the physics major so that you can enter in your junior year and from then you can just cram the rest of the physics major in two years.” Rupesh believed in me so much that he decided to become my mentor. He dedicated his entire summer to mentoring me and tutoring me so every day from 10 in the morning till 9PM at night, we spend time together learning everything from books and I didn’t have time to learn all the theory but it was like Saturday derivatives, Sunday integrals, Monday first three chapters of classical mechanics and so on.
Deborah: At the end of the summer, Rupesh disappeared so that I would really do this on my own. I took the test and I passed. I enter my first physics class was already a junior level class. I remember trying not to burn too many capacitors in the electronics lab. But it was just incredibly challenging. But I managed to graduate with highest honors and Summa Cum Laude with a degree in both philosophy and physics from Brandeis because I persevered so much. That taught me that hard work is what gets people to the end in their path not innate talent. Although there’s obviously some of that was there as well. But the incredible thing about Rupesh is I always wanted to pay him for all that he did for me. I wanted to pay him for his tutoring and mentoring and he said to me that when he was growing up in Darjeeling in India, the tea, there was an old man who used to climb up the mountain and teach him and his sisters the tabla, which is a musical instrument, English and math.
Deborah: The old man said, “No, you can’t pay me. The only way you could ever pay me back is if you do this with someone else in the world.” Kirill that’s how my mission in life began to pass the torch of knowledge so to speak and become that person that inspires other minority students, especially women who like myself feel attracted to STEM, but who for some reason, whether it be financial or social feel that they cannot achieve their dreams. That’s when it all started. I then went back to Mexico and realized that I wanted to continue to study physics. It wasn’t enough to get my BA and I went back to an environment that told me, “Okay, you did your crazy thing, now it’s time to settle down. You got your degree. Now it’s time to get married and have a family and stay home and possibly not get a job in physics.” I was just really sad because I wanted to pursue more, my curiosity was just infinite.
Deborah: I decided, again to apply to universities in the US so that I could pursue my PhD. One day I was studying a master’s degree at the public university UNAM in physics, at UNAM. I went to my advisors office and I told them, “I think I want to move to the US back again because the kind of physics that I want to pursue my PhD in they have better resources there. It’s experimental and the labs are able to purchase incredible equipment and there’s some people doing amazing research.” I applied and he said, “Okay, where did you apply?” I said, “Well, there’s a guy”, I just said it like that. “There’s a guy at Stanford”, I always heard about this university called Stanford that my friends applied to. There’s a guy there by the name of Steve Chu that is doing biophysics. For the first time he’s analyzing DNA, but one single strand of DNA by hatching the DNA strand in an optical tweezer basically, you put in electric field so that you can stretch the strand and see how it behaves and it’s fascinating research. I wrote to him directly. My advisors’ jaw dropped.
Deborah: He said, “What? Steve Chu?” I said, “Yes. Why?” He said, “You don’t realize he just won the Nobel Prize?” I was like, I’m not shy by any means. I probably wouldn’t have written such a casual email like, “Hey, I would love to work in your group.” I was directly accepted to work in Steve Chus’ group and it was the opportunity of a lifetime. Of course, I said yes, took the flight, went to Stanford and it was incredibly challenging. Nothing came easy to me, but after six years, I was told that I became the first Mexican woman to get a PhD in physics from Stanford.
Kirill: Congrats. That’s amazing. This episode is brought to you by SuperDataScience, our online membership platform for learning data science at any level. We’ve got over two and a half thousand video tutorials over 200 hours of content and 30 plus courses with new courses being added on average once per month. All of that and more you get as part of your membership at SuperDataScience. Don’t hold off, sign up today at www.www.superdatascience.com. Secure your membership and take your data science skills to the next level. What an amazing story from the Brandeis University to Stanford and all the challenges and I love the way you described it actually helps understand what somebody who’s interested, a female student who’s interested in physics is actually going through emotionally and experiencing.
Kirill: It’s really hard for me as a man to just perceive it but from your story I get a better sense for it and of course it’s terrible like all the bias that you have to face along the way. Now you have this mission of helping women get into STEM and be successful. Tell us a bit about that. How do you help women get into STEM? What are some of the successes that you can share that you’ve had in this area?
Deborah: I have one success that I’m especially proud of. Her name is Graciela Garcia. I do a lot of workshops from science fairs to helping initiatives like Technovation challenge, where I help mentor some of the girls and develop Technovation challenge is an app creating competition that takes girls from underserved backgrounds and forms teams that get mentorship in three areas, how to program an app that can solve a problem in the community, two, how to create a business model around the app, and three, how to pitch that app to real business investors so that they can have a business model and make money out of the app. I once went to a school to talk about Technovation challenge here in New York, there was this really curious girl by the name of Graciela. She was very young, and she was a little shy, but at the end of my workshop at my talk she came up to me and you could tell that she wanted to do more and wanted to learn more about programming, and STEM and all that.
Deborah: However, she was undocumented in the US and so her mother was very weary of sending Graciela to summer camps or after school programs because of their immigration status. Graciela had not really tested her skills and her curiosity at pretty much nowhere and I decided to take her under my wing. I said you have to come with me and I’m going to mentor you and we’re going to have you succeed. I mentor her and within one year her team ended up winning the final international competition of Technovation challenge. She became after the that their app is called Arrive which helps notify parents when kids have arrived at their school.
Kirill: I’ve seen that app. A commercial for that app. That’s really cool.
Deborah: Yes, because Google is the one that sponsor Technovation challenge and not only do they sponsor with the prize money but they also help with the development of the app a continuing development and so she founded the Arrive app with her team and now she’s graduating from University of Pennsylvania, she’s the first person in her entire family to go to college. This is an amazing score with a degree in computer science. We’ve done TV episodes together, we were interviewed by CNN. We’ve given talks together and she’s just this wonderful, eloquent woman who is succeeding greatly and she would be a great addition to any tech team at any company.
Deborah: That’s success I’m very proud of but I continue to do right now in July, I’m offering a summer camp for girls ages 17 to 20 who are in college or right before going to college on how to succeed in STEM. We’ll do a little bit of coding as well as learning about imposter complex, how to be confident in certain situations in male dominated environments, etc. That’s one thing I’m doing and I keep giving talks and running workshops all around the world for young women.
Kirill: Amazing. How can people find more about this workshop? Sorry about the summer camp?
Deborah: Thanks, Kirill. That’s on my website. My website is sciencewithdebbie.com. That is science S-C-I-E-N-C-E with Debbie is spelled D-E-B-B-I-E dot com. I post everything there my public speaking engagements, my workshops, I posted some resources about how to teach kids of different ages coding and science-based approaches during our challenging COVID lockdown.
Kirill: Fantastic. Really cool, sciencewithdebbie.com. Check it out. If you’d like to learn more about the summer camp or workshops. Tell us a bit more please about inspiring women in STEM versus how can companies take a next step? For instance, what do you do in these workshops? How do you run a workshop? I’m assuming you come into a company or now maybe virtually and you explain how to create an environment where women will feel comfortable and inspired, empowered to be successful in STEM. Tell us a bit about that. Maybe you can share some insights on this podcast for the business owners who are listening who might want to implement a few of these tips that you can share in their business already today or tomorrow.
Deborah: Absolutely. My expertise is in not only helping individual women but in creating policies at company levels that create an environment in which young women are more able to be promoted, feel comfortable, come back after maternity leave and use their different various skills to create products that have a diverse statistical pool of opinions and feedback around them especially data products. I help companies with all of these functions. One piece of literature that I highly recommend for C level executives and people even in HR positions at companies is Carol Dwecks’ mindset book. Carol Dweck is a psychologist at Stanford, and she basically came up with a framework that allows people to see what are the tendencies in women’s bias education when it comes to STEM versus men’s education.
Deborah: For example, she talks a lot about the fixed versus the growth mindset. A fixed mindset is someone and let’s already say this that typically she says women, young girls are educated with a fixed mindset, typically. A fixed mindset tells you that you have a specific amount of intelligence that it cannot be stretched, it cannot grow. It’s fixed. You’re either a B grade person or an A plus or A minus and that’s where you’re going to be for the rest of your life. You better just know it. It’s typically told from the perspective of people who have to pick a career based on their existing skills and not on their hopes to learn new skills. It creates people especially young women who are afraid to risk getting a bad grade because they always have to look like they are experts or they’re trying their best. They’re very conscious of not failing.
Deborah: Whereas men’s education, a young man is exactly the opposite. It’s the growth mindset which basically sets a framework of failure is cool. You should fail because that’s the way to learn. There isn’t a fixed amount of intelligence you can always become an expert at anything you want by failing and trying again. It creates the mindset of people who are going to sign up for experiment things that they’re not already good at like women. More men want to experience say an in introductory course in physics because even though they have not tried it before, they’re not afraid to get a lesser grade or do not feel they’re experts at it whereas women have been conditioned to have to look perfect. Like Reshma Saujani says with her book, educate girls who are brave, not perfect because we are conditioned to appear perfect and that’s part of the problem.
Deborah: Why this is so important, Kirill is because in science failure and in technology is super important. Many results come out of failures. Many PhDs, many discoveries are negative discoveries meaning we discover that there’s no magnetic monopole for example, and people spend their entire research careers researching hoping to find the magnetic monopole like the electric.
Kirill: I was doing that too back in my bachelor’s, I was trying to build a magnetic monopole myself. I know exactly what you’re talking about.
Deborah: I remember Blas Cabrera at Stanford he at one point, they thought they got a result and they discover the magnetic monopole contradicted and that was not it. But that failure he could consider I never found it is actually a discovery that advances science greatly. Data science, software engineering, it’s all about trying and trying again. It’s not about people who inherently know all the answers, quite the contrary. It’s people who are very curious and are not afraid of failure. Those are the ones that succeed. I believe that if you and there are policies that support this game structure of people trying different things and not being punished for failure. If you’re able to do that and of course I give very specific ways of creating policies where women are able to take more risks without being scared of not earning a promotion or not getting certain places at the company if you do it in such a way that women feel they can be built up and supported by the company, then it creates a very successful environment.
Deborah: I give specific advice to companies on how to create policies based on these framework of the growth mindset to allow for women to take more risks and participate and be stakeholders in projects that are more cutting edge and more the R&D stage and part of a company and that’s the framework that I’ve used to a great amount of success.
Kirill: Okay. Could you give us one tip of how somebody will say I’m an executive of the company, what’s one thing I could already do today that will encourage this framework basically show women that they’re encouraged to try and fail then keep growing and experimenting and progressing their career?
Deborah: Sure. Especially in data science, so for example, in software engineering because it’s an older profession it’s been there for a while, things and the tools that you use are very well-defined. If after two weeks for example, you come to your manager, as a software engineer, let’s say your task was to create an e-commerce site and you know the tools to use and the stack and you develop if you could after two weeks to your manager and you say I have nothing to show for the past two weeks then clearly that’s considered a failure. But in data science, it’s very different because if you come to your manager after two weeks of experimenting and doing what’s called exploratory data analysis, if you have nothing to show because your algorithms haven’t work, you haven’t found the right tools or you haven’t found the right parameters that’s not a failure.
Deborah: What I recommend is that each manager be educated in data literacy and in how to promote different styles of management to know that for example, there are projects in which like I said, an experimental approach that’s prone to failure is actually a good thing. To promote women to those positions where they get to experiment, but they don’t get a manager who is disappointed because there’s apparently nothing to show yet and promote women and have the management skill and style that lets them know that their experimenting is totally okay and that they’re going to be supported in coming up with the answers. But I think that’s the number one thing that is the low hanging fruit because a lot of women are because of what I explained before are very afraid of negative feedback and what are people going to say I have to appear as I’m producing a lot.
Deborah: I have to work twice as hard as men to be respected and whatnot and so immediately allowing them to work in projects where failure is not considered failure. Not having a result is not considered a failure, but part of the project, exercising that muscle doing it over and over again allows women to grow and feel comfortable in this experimenting style of work which is very much needed in quantitative fields.
Kirill: Fantastic. Thank you. That’s a great tip. From your experience, you’re a senior data scientist at Metis. We’ll talk more about that in a bit. From where your experience doing workshops and trainings in different companies on topics of data science, what have you noticed is the problem of women not being as prevalent in the STEM fields? Is that getting better in data science over time? Have you seen an improvement over the past couple of years?
Deborah: I definitely have seen an improvement from when I was growing up. There’re workshops and institutions that are specifically helping women, girls who code, black girls who code, women who code, R ladies, Python ladies, there’s tons of support in the community. Having said that, we still experience that at executive level positions it stops. There are very few women being represented. What happens is that this is not something pertaining exclusively to women, but a lot of technical positions, what’s called an IC, an individual contributor people don’t know how to grow and promote those positions. You go stale after arriving at a senior data scientist for example and from there a lot of companies make you manage a team or going to management and completely abandon the field of contributing quantitatively to solving a problem.
Deborah: I think that at executive levels you need to give women more responsibility and have a path of growth and of leadership that don’t necessarily diverge into management. That is career positions where women can still be individual contributors and do technical work because they are good at it and they like doing it and they contribute to results that are quantitative in nature. Create that path but open up new promotion strategies and new ways of contributing to that ecosystem whether it’s in public speaking and being promoted by the companies in evangelizing data products, etc. I think that’s where we’re still failing. Yet one more thing I want to say about this is that we’re also failing young women. Because in this race to make more women learn how to code and more minorities become proficient at coding and technical knowledge, we are just teaching them to code for coding sake.
Deborah: Meaning a lot of schools are like, use Google’s one hour of code so we can all put a check mark in. Our school is teaching how to code. But I had an experience that was actually pretty sad. I went to the Museum of Natural History here in New York, and there were high school girls learning SQL and how to manipulate databases that contained knowledge about the museum. Animals. For example, there was a group of high school girls that were proficient that had been analyzing the museum’s turtles, data about the turtles, there was another group analyzing the birds and other groups were analyzing whales and ocean creature data. I went to the group of girls that was way more proficient at SQL than I had ever been at that age. They were very smart girls very curious. I just said, “Let me look at the data that you have.” I asked a very simple question.
Deborah: There was a column there about the turtles. The column was weight. They had numbers in the column like 130 and 300, and 250. I asked them, “You’ve seen the turtles in the museum?” They said, “Yes.” I said, “How big are they?” They’re like, “They’re tiny. Actually, they fit in the palm of my hand.” I said, “Wow. Is that weight in pounds? What are the units? Is it in kilograms? Is it in grams? What is it?” They’re all stunned because they had not stopped to think about the data that they were manipulating with ease for the past three months. Finally, one girl raises her hand and she says, “I’m pretty sure it’s pounds.” I said, “Wow, that’s pretty [inaudible 00:43:22] because I weigh about 130 pounds or so. Look at my size. You’re telling me a little tiny turtle that fits in the palm of your hand weighs 300 pounds. That doesn’t make any sense.”
Deborah: Eventually we discover and we went back that the units were grams but this to me signaled a wider problem that is pervasive. That is in our race to very quickly get people to program and learn the basics of coding, we’re forgetting what we’re coding for. We’re forgetting that coding is just a tool. But it has to be a tool for something else. The most important thing in quantitative fields like data science is understanding and having critical thinking skills, having people who are able to solve problems. I think remembering that and giving more workshops about critical thinking not just about the skills and the tools is something that we need to do immediately at the high school and middle school level.
Kirill: How can a individual data scientists foster creative thinking skills in themselves?
Deborah: I give workshops on how to foster critical thinking skills and I think the answer is in what little kids do in asking lots of questions about the data, about the sources of the data, what are the goals of analyzing this data? What are the actions that a company is going to take based on the insights that come out of analyzing this data? Who’s really a stakeholder and for what reason? Is it an ethical approach that I’m using or am I biasing my, well we’re always going to use statistical samples that are biased in some way or another but making an effort to make it less biased towards certain groups. I’m I using this for a certain purpose when it could be used for different purposes? There are a set of questions that should be asked of the data of the company of the agenda that the C level executives for example in a company have that allows people to think clearly about what the goals are for the different projects that get thrown at them.
Deborah: They can become leaders and saying we’re not doing this right we’re actually forgetting and I’ll give you concrete examples. There was an algorithm in Facebook that failed because it was an image recognition algorithm that recognizes people in pictures. It was not trained with enough black people in the beginning and so it would fail miserably with those groups. It was clearly very biased so that’s something that should not happen in this day and age. There’s another product, the soap dispensers where you put your hands below them in airports and public restrooms and it dispenses a little bit of soap. It turns out and there are videos on YouTube showing this it did not work well with dark skin because it was calibrated. The laser was calibrated for the comp to recognize the contrast between a white skin hand and the dark background of the equipment.
Deborah: The African American person is trying and trying and it’s not dispensing soap. There’s also the discovery that the airbags for safety that were built for cars, the team there were no women in the team and it was all white males of a certain size and age, doing all the testing. The bags a lot of women were having accidents and the airbag would either not inflate or would cause issues like choking and different issues because their size, the momentum that a typical female body experiences were not taken into account. This is why I think being a critical thinker is how can I include a really wide base of users in my testing. How can I make my algorithm work for different instances? How am I biasing the data? All these questions are super important for critical thinking.
Kirill: Fantastic. Thank you. That’s some great examples, as well as advice on how to be a better critical thinker. I’d like to shift gears a bit and talk about your data science work. You’re the chief data scientist at Metis. You’ve been at Metis for almost five years which is very exciting. Tell us a bit about the work that you do.
Deborah: Sure. As a chief data scientist since Metis’ inception I was in charge of developing curriculum for our boot camps that are intensive workshops or boot camps that take people from very little knowledge of both programming and the mathematics and data science, all the way to becoming proficient and off to get jobs in data science. I checked for what were the new tools that were being used for example, D3 for visualization was no longer being used. What could we teach instead for visualizing graphs, etc. I was in charge of managing the team, day-to-day activities of 16 senior data scientists, I’m no longer in charge of the operational management. I’m now more in charge of the thought leadership. Getting all our instructors to be public speakers to participate in the ecosystem of data science conversations out there.
Deborah: I also worked in developing new programs outside of our bootcamp. I managed the relationship with Dublin Business School for example, we created a machine learning course that was specifically for their masters in analytics program. I talked to various corporate institutions that wanted a corporate training type of offering. This is what I do more now which is more related to taking the data science content and shaping it in such a way that it can be applicable and it can fit different institutions depending on their needs.
Kirill: Fantastic. What would you say is there a golden formula that you have on how to pick the right curriculum for your team or for even yourself if you’re trying to learn data science?
Deborah: I think there’s a whole discovery process because in data science there’s three things and data literacy that I think are important. Data, the product and the techniques. Most people are getting these wrong. A lot of companies have a ton of data and they don’t know what to do with it. Should I hire people to build tools from scratch? Should I use open source tools? Or should I have proprietary tools? If you’re a large company and you’re concerned with auditing and compliance definitely, will probably need more proprietary tools. Then techniques like should I hire an expert in deep learning? Or do I only need someone to automate Python scripts to do this job? To have the three aligned and serve the entire company is what I call true data literacy. That takes a whole conversation with a company.
Deborah: We have figured out that when we go to companies and the people asking us for a workshop or for a particular training are not aligned with the people receiving the training, then all hell breaks loose because the executives are paying for the training but they have no idea what the problems, issues are of the stakeholders that are using the company’s data. These people are the ones that are going to take the training and if they have no idea why they’re taking it and what are the goals, is it faster algorithms? Is it a different data product that I’m supposed to build? There’s a lot of confusion here. Our discovery process initially with every company that we speak to is I would say the most important phase, because if you don’t have the right tools but have the right people, the project will fail. If you don’t have the data, but have the right tools then the project will likely fail and so on. You need to have these three things aligned for a project to really be successful.
Kirill: Amazing. That’s really good advice. For an individual data scientists, the takeaway here is that if your company is designating certain training to you, or you have available training that your whole team is going through or something like that, I always ask the question, is this something you really need? Maybe the person that has assigned this training doesn’t really know the problems that you’re working on.
Deborah: Yes, this is an analogy from physics. Physicists are very proud people. We always pride ourselves in using economy of mathematics. By that I mean if you can write an equation in shorter and shorter ways and in more elegant ways then it’s definitely better. If it has the same explainability and encompasses all the different use cases that you should use the formula for. I think that should be a learning thing for data scientists. If you can analyze these data with a simpler algorithm, why not do that? Many times without critical thinking, we think, let’s kill an ant with a cannon. This is a saying we have in Spanish. We don’t want to do that. We want to use the right tools to solve the right problems. If I’m tasked with solving a data-driven problem, then I should absolutely first do exploratory data analysis and see how I can just do a linear regression and it may work and so on. These are the different things that people should think about when being tasked with analyzing a project.
Kirill: What’s the most common training requests that you get from companies?
Deborah: It varies quite a bit, but I would say it’s more on the simpler. You would think people are asking us for deep learning or quite sophisticated training, the reality is that very few companies are stuck there or see great value in that. I’m talking, obviously the companies that do see great value in deep learning are typically very large, established companies that have already internal teams doing deep learning by no means meant, I know it sounded like I was saying they’re not fruitful algorithms. Quite the contrary they’re giving us incredible visual recognition things but typically those are companies that have ingrained in them a data science practice. The companies that are seeking help from us in training are companies that don’t yet know what kind of team to build.
Deborah: Should I upgrade the skills of my software engineers? For example. Should I make tools like Tableau available to my HR and more junior level analysts so that they can also help interpret the data and get insights? I always say that the best tools are the ones that people don’t even realize that they’re using these very sophisticated tools but it’s giving them the data that they need at a time that they need it and in a way that’s easily usable for them.
Deborah: That’s the type of help that were asked like, “Should I automate a Python script so that my analysts don’t have to spend four hours loading the stock market’s data onto Excel?” That’s one example. Another one could be how can I make the process of analyzing product recommendation for my e-commerce site faster and more efficient? What recommendation algorithms can I build? How can I test different parameters in a way that’s efficient and it’s not going to consume all the time of my data science team. Things like where it’s at a more simpler level that help me with my well-tested five year running, deep learning algorithm. Tweak a little bit to make it even more efficient.
Kirill: Got you. Then you have all this content inside Metis. Then you find all right, what’s the right content for this client structure the learning program and off you go.
Deborah: Yes, not only do we have great intro courses, Intro to Python and the mathematics for data science, we have a data literacy coursework. We also are able to customize different products depending on the needs of the company. We’re able to do that because our expertise lies in many different areas. Our instructors come from various backgrounds, a lot of them have PhDs in certain specific areas. We at this moment are still able to offer pretty tailored data processes for companies that require something off their processes to do.
Kirill: What’s the percentage in demand for either in-person or virtual instructor led training versus on demand training? What is more popular? Whether it’s led by an instructor, whether it’s virtual or in-person, or whether it’s pre-recorded videos that the company can assigned to their employees, which is more popular?
Deborah: I would say definitely more on demand one. The asynchronous content and just giving them videos and explaining the skills and the task to do is successful to a certain extent but over the long run it’s a little bit like Coursera and these online courses. If you don’t assess the level of skills that was gained by the training, then it’s very difficult to know if your investment had a good ROI. Had a good return and that people actually ended up doing it. You need to have some contact with the team that’s being trained either in person or virtual but definitely that mentorship and that continuous assessment of the increase in their specific skills is very useful and I would say essential for a true data literacy progress in a company.
Kirill: Wonderful, thank you and you have a really cool case study which is publicly available. I think it’s the CMA. Would you mind sharing that with us please?
Deborah: Sure. A CMA is a boutique consulting firm and I just shared an article on LinkedIn. You can search with my name Deborah Berebichez and you can read the article on data literacy and basically I talk about the wrong ways of doing data literacy at a company and the right ways and CMA is an example of the right way to do it. Because basically, they first had internal talks where the whole company, C level executives, HR people, analysts, junior people were aligned in the goals of having a data science practice. The wrong way, done by many companies is like I said before, bringing in a team of data science trainers to a situation where there’s no communication and no alignment between the people who purchase the training and the people who are receiving the training.
Deborah: CMA was very well-aligned. The stakeholders of the data were at each level and in each branch of the company. They all knew why they were taking the data science training, they all knew what they were expected to learn from it and what the goals were in the end, either saving time saving money for the company or creating more efficient algorithms and whatnot for the different areas that they do consulting on. I would say, that was a great example of how you need internal buy-in before you go ahead and say, “Hey, let’s get some people to train us on how to do data science.” Not really having talked in deep terms with the teams that are actually using the data. The other thing that was successful is that at every level, people had the right skills and the right tools to gain insights that were useful for their particular job function.
Deborah: That’s really good because you’re not going to give say an HR person that is not attracted by mathematics and statistics and programming, you’re not going to give them open source code to be working with, but you may give them Tableau or some other platform that simply outputs the right insights that they need at the right time so that they can do their job better. At the same time you are going to give the sophisticated data scientists that are doing your R&D at the company, you are going to give them the full stack of tools to be able to advance the efficiency and expertise of the company.
Kirill: That’s really cool. What results did CMA experience after this these trainings?
Deborah: Well, from what we’ve heard from them is that they were able to automate certain things that previously took them a long time. They actually had more integration across teams, because now they spoke data-driven language in common so they were able to have metrics of success that were predefined and now everyone was able to know what those typically jargon words mean. What’s a turnover? What’s a recommendation rate of success and things like that. Now that people had the right words and the right tools obviously being presented on screen in different ways and at different levels now they were able to communicate across teams much more effectively, efficiently.
Kirill: Critical. Very exciting. Thank you for sharing that. That’s a great example of a success and also contrast to when people don’t know why they’re learning something and why they’re doing certain things. To finish off, I wanted to ask you about how you see the future of data science? From what you’ve seen in this industry and through your personal experiences, what do you think we should be looking forward to or anticipating in data science in the next, let’s say five years?
Deborah: I think this field has been trying for a very long time to define itself, unsuccessfully. You as a data scientist, let’s start with software engineering, you as a software engineer will probably end up doing something quite similar. If you get hired by two different companies. Yes, the details may change and the tools may change but it’s pretty well-defined what your skills are and what your tasks should be. However, a data scientist is still a pretty undefined field. That is if you get hired by Facebook as a data scientist and then graduate on to work at a hedge fund to do data science, what the title data scientist means at these two very different companies is something completely different. You’re going to end up having different goals and different tools and different kind of data to use that requires different expertise and very different skills.
Deborah: What companies are realizing is that they need to really learn into how to categorize the skills of data scientists, maybe somebody certified in classification algorithms another person does deep learning. What I expect in the future is for more well-defined skill sets and applications to the specific algorithm in the field. You’re going to be able to have commoditized products very much like Tableau and other products that are going to be much more widely available for people who are not data scientists, per se the technical people. Then you’re going to have special fields of data scientists that are going to continuously develop tools that are more advanced for various other skill sets and purposes at a company and those are going to be different silos that are going to be very separate and very different one from the other. When you see job postings, you no longer going to see senior data scientists. You’re going to see senior data scientists focused or with experience in NLP for example, Natural Language Processing and that’s going to be a very specific skill set.
Deborah: People are going to also specify by the field, what’s called the expertise or being well-versed in a certain topic. People who have worked with health data have acquired certain knowledge about the parameters and the way they look at the data that’s quite different than people who are used to working with numbers in Wall Street and at hedge funds. You’re going to see many more silos and a wider acceptance of using the tools that are available for everyday data-driven insights.
Kirill: Interesting. In terms of silos, it’s going to be the first example that comes to my mind is marketing for instance marketing, a marketing specialist might be with experience or expertise in affiliate marketing versus email marketing versus paid advertising versus content and sales. There’s four different areas and with a background in, as you say, retail or SAS or B2B or whatever else. My question is, I could see how that’s a very valid point because maybe 20, 30 years ago, that wasn’t that delineation, those silos weren’t there. My question is for data science and for data scientists are silos good or bad?
Deborah: Well, they’re good because they define the boundaries and they have more clarity in aligning the goals of that team and within a company of that silo, in that sense they are good because a lot of data scientists feel frustrated that go on to do something really sophisticated but they lack the domain expertise in something else and their insights are not being used. A lot of work is being wasted. They’re bad in the sense that interdisciplinarity and cross-referencing various types of algorithms and domain knowledge, applying something from the insurance field into healthcare it could be incredible. Silos are always bad for those aha moments. What I think is going to happen is the more everyday level, the silos are going to be more defined but at the R&D level, there’s going to be a separate lab so to speak where people are going to be testing new things, cross-pollinating different areas and different domain expertise as they’re seeing the algorithms that are applicable to various fields and whatnot.
Deborah: In fact, I predict that with the ubiquity of Internet of Things that is of sensors, and the data that’s coming from sensors, we’re producing more data in two days than we used to produce in analyzing the sky or the Large Hadron Collider Subatomic Collisions in years. All these data that’s coming from sensors we’re going to also learn new tools of how to refine the way in which we select the information that is useful from the one that is not useful. That’s going to bring a more engineering automation of tools that are going to continuously select information that with practice and with years of experience we already know we need in order to gain insights and companies are not going to store like now all the data just because we don’t know when we might need it. That’s going to be discarded and then the relevant data is going to be put into production lines.
Kirill: That’s a really fantastic description. That was probably one of the most vivid descriptions of the future of data science I’ve heard on this podcast. Thank you so much is extremely exciting. On that note, we’ve come to an end slowly of this episode. Thank you so much for being here today. I would like to ask where’s the best places for our listeners to find you and follow your career or maybe engage with some of the work that you’re doing, whether it’s in data science or in inspiring women in STEM?
Deborah: Thank you Kirill. I always love when people reach out so please do so. You can be curious about anything and I always love to converse and form intellectual friendships. It’s great when I hear from people who are curious. I co-host a TV show with the discovery channel that’s called Outrageous Acts of Science. You can watch all of our past episodes of 11 seasons on iTunes, Amazon Prime, and I believe YouTube. You can catch up with me directly through following me on Twitter. My handle is Debbie Bere, D-E-B-B-E and then B as in boy, E-R-E. Debbie Bere the first four letters of my last name. I’m on Instagram, I’m on Facebook. I have a public page on Facebook so you can follow all my workshops and what I do and of course in my website I have a direct email that you can use info at sciencewithdebbie.com. You can communicate with me and we can figure out how we can work together.
Kirill: Fantastic. Is it okay to connect on LinkedIn as well?
Deborah: Yes. Sorry. Of course. LinkedIn is wonderful too please connect with me there.
Kirill: Fantastic. Wonderful. Well, thank you very much, Debbie. It’s been a huge pleasure having you on the show. Best of luck with your TV shows and all the amazing things that you do especially the mentoring of women in STEM. It’s been a great discussion. Thank you.
Deborah: Thank you, Kirill. Have a great rest of the day.
Kirill: There you have it, everybody. Thank you so much for being here for spending this hour with us. I hope you enjoyed this podcast as much as I did and got lots of valuable takeaways. My favorite part of this podcast was the future of data science and what Dr. Deborah Berebichez said about silos and how they will eventually form in the industry as it matures. The benefits and the drawbacks of having these silos. I think it’s something new that I haven’t heard before on this podcast. A very vivid description of the future that’s coming and something definitely to look out for. We’re quite lucky to be in this industry when the silos are not formed and experience it this way, and then also experience it once the silos have been formed that’ll probably be the next decade or so.
Kirill: In addition, of course, the first half of this episode was extremely important in terms of helping promote equality and inclusiveness in the space of data science and in general fields in science, technology, engineering, mathematics. I hope it was useful to become more aware of what’s going on in this space. Definitely Deborah’s personal story was very inspiring to hear. I think we all should do as much as we can to make this space as inclusive and as encouraging and inspiring to absolutely everybody who wants to be part of it. On that note, as usual, you can get all the show notes for this episode at www.superdatascience.com/377. That’s www.superdatascience.com/377. There you can find the transcript for this episode plus any materials we mentioned on this show any links URLs or where to connect with Dr. Berebichez and all the other amazing things that you may have heard on the episode and you’re looking to find they’ll be there.
Kirill: One final thing I’d like to ask of you today is if you know somebody who’s a female in data science or who’s just part of a minority in the space of data science, please send them this episode. This is a very exciting story, a inspiring story of a person who went through all these biases and challenges and troubles and still was able to be successful in STEM and build a career and help inspire others and pass on the torch. This is your opportunity to help inspire somebody else. If you know someone, then send them this episode very easy to share, just send them the link www.superdatascience.com/377. You might actually change somebody’s life by doing that. On that note, thank you so much for being here today and sharing this time with us. I look forward to seeing you back here next time. Until then, happy analyzing.
