MLOps Zoomcamp: Free MLOps course. Register here!


Transitioning from Project Management to Data Science

Season 3, episode 1 of the DataTalks.Club podcast with Ksenia Legostay

We talked about:

  • Ksenia’s background
  • Data analytics vs data science
  • Skills needed for data analytics and data science
  • Benefits of getting a masters degree
  • Useful online courses
  • How project management background can be helpful for the career transition
  • Which skills do PMs need to become data analysts?
  • Going from working with spreadsheets to working with python
  • Kaggle
  • Productionizing machine learning models
  • Getting experience while studying
  • Looking for a job
  • Gap between theory and practice
  • Learning plan for transitioning
  • Last tips and getting involved in projects


Did you like this episode? Check other episodes of the podcast, and register for new events.


Alexey: This week we will talk about transitioning from project management to data science. We have a special guest today — Ksenia. Ksenia is currently working as a data science manager at momox. After four years working as a project manager, she turned her career towards data science. She spent three years researching fraud and anomaly detection techniques and she earned a degree in data analytics. Welcome, Ksenia. (2:24)

Ksenia: Hello, everyone. It's a pleasure to be here. Thank you. (2:57)

Ksenia’s background

Alexey: Before we start, before we go to our main topic, can you tell us about your career journey? (3:00)

Ksenia: In 2013, I got fundamental education in math, in Russia, in a quite prominent university. Then I got another management degree in 2014. Then I spent four years as Alexey mentioned in project management. I worked and managed projects in e-services, governmental services, traveling, telecom, and other fields. Then I acquired another master's degree in Information System Management. In parallel, I was doing research for three years and I will explain about it later on in further details. I'm working as a data scientist since 2019. Already for three years. Currently I'm working in momox and I shifted my focus from fraud detection to marketing. Now I'm applying data science techniques in marketing. It's quite entertaining and I really like this field so far. There are a lot of optimization challenges, a lot of things to predict and improve. It's really a nice place and position to work. (3:09)

Alexey: Data science attracts many people from quite different backgrounds. For example, I am a software engineer. Many people who are working in data analytics also go into data science. And sometimes project managers like yourself get interested in this. I’m curious, how did it happen for you? When you realize that machine learning is such an interesting field? (4:35)

Ksenia: This question that I have heard so many times. I will rephrase it. Before going into machine learning, I realized that data analysis is super interesting for me. When I worked in project management, I worked closely to development. My work was associated with decision making. I was curious, every time when I make some decision, is it right or not? Why should I decide it? How can I decide it? I don't know. At those times concepts like “know your customers” or principled in marketing, and customer centric way of thinking have been developed and integrated in businesses. People started to learn what customers want. And they shared a lot of data that companies have towards their products. You can analyze it and it would be not smart to not use this data. In those times, I realized that I want to make a decision, which would improve the product, but based on the customer's preferences. Because it's not me who decides what direction the products should go. But it's the customers who actually drive this business and who consumes this product. I think it's super important to analyze data and to understand the sense out of this — what your customer wants. That's how I got into data analysis and that's why I became interested in this field. (5:16)

Alexey: You worked as a project manager, and you needed to make all the decisions. Usually, to make these decisions, you would rely on data. You realized that analyzing data is such an interesting thing. Then you got into data analytics more for supporting this decision-making process. Then, after that, you transitioned from data analytics to more machine learning and data science. (6:54)

Data analytics vs data science

Ksenia: I would say this is the next step. When you analyze data and you feel comfortable. This is when you learn some packages to analyze data, to visualize data, to deliver your findings and also interpreted on the business level. Then you start becoming interested in “What’s next?” I derive this knowledge, but what will my customer do next? I personally think that data analysis is more focused on historical data and deriving a sense of what your customer did or how your business went so far. Data science is forecasting the future so you try to predict what's going on. I think that you probably can't go to data science without transitions into data analytics. Data analysis is the main part of any data science project. Because you have to analyze your data — just to build your hypothesis on top of what you observed. (7:30)

Alexey: We've had a question, “what is the difference between data analytics and data science?” And I think you just covered that. If somebody decides to study data science, how do they do it? How did it happen for you? You realized that there’s a lot of potential in data. That I need to analyze this data, I need to understand what's going on? How did you decide which things to study to be able to do this? (8:33)

Which things to study?

Ksenia: Firstly, I tried to search through the internet by myself and just find some courses on Udemy, and Coursera, and so on. But soon, if you do the same, you'll figure out that the amount of knowledge is immense. You’re lost where to start. I felt the same when I started to approach this topic. My benefit was that I already got familiar with math, statistics, probability theory, and everything fundamental that you need for data science. I think that you have to come out of your background. If you try to shift your career from backend development, you're already experienced in coding, engineering stuff and so on. So you probably want to learn some packages for data analysis and also improve your statistical knowledge. If you come from data analysis, it’s another pool of information that you have to пуе before. So you have to understand your strengths. Then realize what is lacking for you in your skill set, and then add it up to your profile. For me, it was like, I’m really fond of fundamental things and I already got two degrees and I decided “Okay why not get a third one?” That's why I choose my path. (9:11)

Alexey: If I understood correctly, you mentioned that you already had a mathematical background. You already knew statistics. Is it something you needed to use as a project manager? Or this is something that you learn during your studies? (10:52)

Ksenia: Yeah, I was just lucky that I got math as my first degree. I have to say that at 17 years old, I definitely knew what I wanted to do for my life and knew what to learn and so on. My math was chosen and I learned it. I probably did not use it so extensively in my project management career. But I'm really grateful that I had it in my background, because it helped me a lot when I got into this data analytics degree. It was super useful. I knew all this stuff and the main thing was just to remember it from the bottom of my head. (11:10)

Alexey: So you studied mathematics during your bachelor's degree. You worked as a project manager. Then you got into data analytics, you started to learn all these things. And then you realized and this stuff that they learned at university was actually really useful. (11:55)

Ksenia: Yeah, but it wasn't like this. It was like “Okay, I know math and it's my strengths. How can I improve my profile based on this?” I don't want to waste what I already have. Data analysis and machine learning was based on what I already have. I want to extend my knowledge. So it wasn't a spontaneous decision. Of course, I realized that I will need math for this and “oh, okay, I know it already”. (12:14)

Alexey: I guess math was not the only thing you needed. In addition to mathematics, what are the other things you needed? (12:49)

Ksenia: You definitely need engineering skills and ability to code. It's super important and we will cover this later on. Another thing, you need to analyze data. Mathematics is just a theory. But when it comes to practice, you have to get a feeling of why you need it, and how you can apply it. Another thing that I find quite important is domain expertise. When you work as a data scientist, you're not working in space, in the middle of nowhere. You're working on specific fields. You have to study it, and it will be nice if you know business KPIs, if you know what you want to achieve in business terms and so on. This is what definitely has to be learned. (13:00)

Benefits of getting a masters degree

Alexey: You also mentioned you tried to pick up these skills online by trying to do courses on Udemy, Udacity. But it was just too much information, too many courses, it was too difficult to choose what to learn, and you decided to get masters for this reason. (14:00)

Ksenia: I think that these people, they're professionals in education. They definitely should know how to teach people. The main reasons that I can summarize why it's worth to go to master’s degree. First of all, it's deep enough. From university, you can have the deepest knowledge you can find. In books by yourself, but it would be difficult. More difficult than just going through a lecture — get this feeling and understand the concept and so on, from people that you can ask. Deep enough, well structured. People who construct this study plan for you, they thought before what should be covered in what order. (14:27)

Ksenia: When you start to trace some websites and internet, you can be lost in understanding what you should study before and next. Your learning plan should be sequential. And universities should know how to do this. Then you have to experience an increasing level of complexity. If you just straightaway dive into neural networks and deep learning, you probably can feel overwhelmed or even demotivated to proceed. It's quite difficult to start with. That's another benefit of going to university because it's structured for you at this increasing pace. Then feedback that I already mentioned. You have to do some exercises, you can ask for feedback, and so on. Another thing that I find really cool here in Germany. During my study, I got a position in a research laboratory and I spent three years during my study as a researcher. I was lucky that I studied something in a lecture, and I immediately applied it in practice. It was so beneficial, because if you learn something new, and you don't apply it straightaway, it will fade out of your head. It's a pity that you have to learn it again and again in the future. This is why I chose a degree.

Alexey: What things did you study? I understood the benefits. It's very hard to argue with these benefits. I'm wondering what concrete things you studied at university. What were the most useful things — what subjects or courses? (16:59)

Ksenia: Because I knew that I wanted to go towards a data science career. At university, there was a lot of freedom to choose. I have chosen the hardest courses. I was really interested in it. I choose machine learning classes — supervised and not unsupervised. Then artificial intelligent classes. They're quite different. They speak about the same problems, but from different angles. Then multivariate analysis, statistics, time series analysis, graph theory, networks analysis, and all this mathematical stuff. I felt that I knew math and probably I can tackle these courses. (17:18)

Alexey: It was quite focused on data science, because you could select the courses you wanted. You knew what you wanted to study and you picked up courses that you thought would be super useful for your career, like machine learning, artificial intelligence, statistics. (18:06)

Ksenia: From my previous experience as a project manager I observed the problems. I know what business needs — or at least I have an impression that I know it. I tried to base my choice of courses on what I observed in the business. It is also another benefit to go to a master program after having some experience. Because you understand what you will be resolving later on at work and why you have to choose this or that. (18:25)

Useful online courses

Alexey: I can already see how your background as a project manager was useful. It feels like it was very carefully planned — the way you approached your studies. You knew what you wanted to do, and you just went for it. That's really cool. I know we already talked about online courses. You had a chance to compare these online courses and the courses you took at university. What is your opinion, how they're different and what worked better for you? (18:55)

Ksenia: I can even give advice to people who are interested in choosing between having a nanodegree degree online and go into university. It depends on your purpose. If you want to pick up some knowledge about a specific package to analyze or visualize data, it's enough to go to code academy and pick the first course that explains the semantics, gives some examples. And you learn it immediately in one or two evenings. But if you really want to understand the deep nature of data science concepts, I can recommend some nice online courses that helped me. I didn't do it during my university. During the study I took a lot of classes and also I worked as a researcher and I had enough practice and knowledge to consume. It was a full learning process. (19:36)

Ksenia: But when I graduated, I realized that sometimes when you do one thing, you might forget something else. You have to refresh your knowledge. I do some classes on machine learning every year just to refresh my knowledge about everything. Because usually at work you don't use everything that you learned. But it's still important to have it in mind. I personally find this fundamental online class and what I could recommend. There is OpenDataScience community and they provide once a year this online course designed by Yury Kashnitsky. It’s a well-structured course that gives you a deep understanding of algorithms and also provides a lot of practice. You can communicate with tutors and also get feedback. You also communicate with peers and learn from them. I personally really like it and guys did nice courses completely for free.

Alexey: I'm googling it right now. It's called “”. I also took that course and it’s amazing. I also took this course when I was already working as a data scientist. When it comes to this understanding of different algorithms I think this is a really great course. (22:00)

Ksenia: I repeated your experience unconsciously. (22:26)

How project management background can be helpful for the career transition

Alexey: I already mentioned that your background as a project manager helped you at least in structuring the way you approached your education, like deciding which courses to take. I think it was really well thought through. Were there any other things from your experience as a project manager that helped you? (22:32)

Ksenia: I will not go in for myself because I'm not that representative here. But I summarized some background that you can transfer from project manager to data science. It's not that obvious and I really like to put some time on this question. Some people think that it's quite obvious to transfer from data engineering or from data analytics to data science. Because you can code, you can analyze data, process it. You already know some query languages and so on. But people underestimate what benefits you can bring to data science coming from project management. (23:07)

Ksenia: This benefit is definitely in soft skills. You are experienced in planning projects. Any data science project is just a project. You have to plan it in advance and you already know how to do this. You have to focus on the big picture, not to restrict yourself to some details, but you have to plan your roadmap. Set some milestones and so on. It's a normal project that you have to set up and you have to decompose it. You have to schedule it. You have to plan it — and you already know how to do it. Usually it’s done by some lead or manager. In data science, being a data science manager, it's the best thing that you can do. Because you know both things from your perspective. You also can define success measures. It’s quite important and really difficult. People tend to define it from their point of view. If you're into business, if you stay close to the business decisions, you can identify success in terms of business, revenue, margin, or what you can imagine. This is what board and managers really like. If you work as a data scientist, you can speak with them in the same language, and they can understand you. You are transparent and it's super important — this is what I learn from my practice.

Ksenia: “Speak the same language” — this is what I mentioned. It has already been covered in one of DataTalks podcasts — by Vin Vasishtha. He spoke a lot about the borders of business language. It was quite a nice talk. What I also find useful and borrowed from project management experience is — be business oriented. When you make some data analysis, you extract and derive some sense, some knowledge. You probably find some outliers, and you plot some distributions. Okay. It's nice, but what does it mean? Not for you, as data scientists or statisticians — what does it mean for business? Where do you lose money? Where do you earn it? Why you lose some customers? Why are churn rates so high? It's really important to put it in the business language, and to understand the reason what your distribution shows you in business terms.

Ksenia: Understanding the business really helps you in problem definitions. You can frame your problem before solving it — [to make it] really clear for everyone. It's also entangled with the business understanding. I think that Elena Samuylova mentioned it in her talk “How your machine learning project will fail”. She spoke a lot about problem definition and how it is important. You definitely should understand how business works. You, as a project manager, already have experience to communicate with stakeholders, to be clear, to deliver the right information at the right time, in the right order. This is important also in data science, because you build your model not for building it, but to deliver value to the business and to communicate it right.

Ksenia: The last thing, but not least, is be proactive. When you are a project manager, you have to be proactive. You have to initiate many processes. You have to organize people, organize things, and plan and so on. This pro-activity is super nice, also in data science, because it fills a gap between technical implementation engineering and mathematical concepts to the business. This is where people could improve themselves to get on business level. This pro-activity and curiosity about business purposes, about a business KPIs, will definitely help to bring your model on the next level and estimated in business KPIs. This is what my project management experience brings me to data science.

Alexey: Thanks for mentioning and some of the presentations we had on DataTalks.Club. For those who are listening, or watching, they are on our YouTube channel. You can go to our channel after this talk and you will find them. One was a talk by Vin Vasishtha about monetizing machine learning. And the other one is from Elena — “how your machine learning project will fail”. Elena's talk is one of our most watched talks [in our channel]. Check it out. (29:00)

CRISP-DM for structuring your projects

Alexey: I wanted to ask you. You mentioned that one of the things that helped you most were planning skills. You were able to plan a project because a data science project is a project. Because you have experienced planning projects, you can take this experience and apply it to data science projects. For those who do not have this background in project management, maybe there is a simple course or a book that they can take to get this planning skill? Maybe you have some recommendations for that. (29:38)

Ksenia: Definitely. It's fortunate that it's already developed. I would recommend to use the standard of the industry in data mining — CRISP-DM framework. It's really convenient for data science projects. I also use it. It's nicely structured. You can structure your project based on the stages of the CRISP-DM framework. It's really well designed. it's one way to go. And regardless, if you come from project management or from whatever, it's nice to use in data science. (30:20)

Moving into analytics for PMs and the most useful skills

Alexey: Yeah, even though it's a bit old methodology, it's 20 years old, it's surprising that it still applies to projects these days. Do you say that planning skill was the most useful one? (31:11)

Ksenia: I think that the most useful skill to have, in my situation at least, is being business oriented. Because your business, your stakeholders, expect from you to deliver value. Of course, you have to plan. If you can be transparent, if you speak the same language, if you're oriented to business KPIs, this brings us more success to really deliver value. Not statistical inferences from the data, but transferring them to the business terms. This would be the most important thing for data scientists. It helps. (31:32)

Alexey: We have a question from Sandela. Do you think an experienced project manager would already be a good fit for a junior data analyst role? Or they still would need to learn some things, to pick up some extra skills to be able to work as a data analyst? (32:16)

Ksenia: I think I mentioned it, but I will elaborate more on my response. Becoming a data analyst after a project manager career is exactly the way to go. You see a lot of data. You have to analyze it. If you have to derive some knowledge out of this — and what I would recommend you to start with — if you're already working as a product manager, and you are eager to tend to your career towards that data science, just start to analyze data that you already have. Because you have your data. You have your project. And I am sure that you know how you can improve your decision based on your data. (32:43)

Ksenia: You can start straight away. It will be amazing, because you will be motivated in doing this. Going after your full day of work to home and picking some knowledge from some courses — it's probably not that motivating. Because you're already tired, and you want to relax. It's overwhelming. But if you really try to apply your knowledge and your skills in data analysis on your work, it's a way to go — it's beneficial for your project management tasks, because you will decide better. You will make a better decision based on the data and based on the knowledge that you've derived. Try to apply your knowledge already while being a project manager and try to get into data analysis as soon as possible.

Going from spreadsheets to SQL and Python

Alexey: I assume that the usual technology stack, the tools the data analysts use, is quite different from what the project manager would use. I imagine that for project managers, the tools will be Microsoft Excel or something similar, like spreadsheet software. So, how would you go from working with excel spreadsheets to SQL or even Python and Pandas? (34:18)

Ksenia: I would recommend to start with frameworks that are already developed for data analysis — drag and drop tools. For example, Tableau and Trifacta for data cleaning and data analysis. You don't need to learn any coding language to start with these tools, but you already can analyze your data. You already can understand some summary statistics from this tool. Just put the data there and try to play around with this to plot some distributions and to analyze it. You don't actually need code at the very beginning. (34:48)

Ksenia: As a next step I would definitely recommend to get some courses on Python — probably just a simple one from Udacity or somewhere. And get some basic knowledge about packages to analyze and visualize data. It's not that difficult because Python has a really low threshold to get in and it's easy to learn. That's why it's so common in the community, because it's really not that hard. I definitely recommend to you, as a next step, to pick some Python knowledge and already apply matplotlib charts or something like this. And Pandas.

Joining communities and joining Kaggle

Alexey: What would be the best way to start with coding? One thing you mentioned — take some courses online. Maybe it doesn't really matter which one, just take one course and try to go through it. What I want to ask — okay somebody did this course — so what's next? How can they apply this to actually solving the problem? (36:16)

Ksenia: I started to analyze data through Python. In university I had some classes but it was mostly recommendations and buzzwords, like “Jupyter notebook”, and you figure out by yourself what's going on. It's not like somebody would teach me. Now I would recommend people to really understand where to go. Join communities. It's super important. That's what I learned, and I regret that I didn't do it early. DataTalks community or OpenDataScience community. Communities are really powerful. Don't be shy — in this community, there are a lot of people on different levels. You can ask whatever you have in mind, and people are quite open and willing to help. Start to learn from these people. Start to be curious about things. That's what I would recommend. You will definitely figure out how people analyze data. (36:47)

Ksenia: Other advice that I remember — open an account on Kaggle. That's what also helped me a lot. Observe and study from others. In Kaggle you will find a lot of open notebooks and. You can understand how people analyze data — try to repeat it. Fake it until you make it right: repeat the best techniques, the best practices — what you observed. Try to pick people who achieved a lot on Kaggle — they probably will code some professional functions and so on. Just try to observe what other people do, if you start to learn. It's important.

Alexey: You don't need to compete on Kaggle to actually learn from the site. You can just go to take a look at these notebooks and try to reproduce what others are doing there. That's already a very good start. (38:54)

Ksenia: Yeah, from the very beginning you probably can not compete, if you don't want. But I did and it was quite fun. As I mentioned, in the OpenDataScience community’s course, one or two homeworks was to compete in a Kaggle competition. This was fun because it was a group of people who try to solve the same problem. We share our notebooks and we learn from each other. If you resolve this problem and your peer resolves this problem, you might find his solution interesting. Learn from others, and somebody will learn from your solution. I think it's nice and it's fun. It's really important to stay motivated during the learning process. (39:09)

Productionizing machine learning models

Alexey: Kaggle is fun and you of course can acquire a lot of useful skills there. You can also do this from courses. Then you can do a couple of courses about Python, a couple of courses about analytics. But at work — if you work as a data scientist — it's not enough. You also need to have some other skills — for example, once you train the model, you need to be able to put it in Flask or Docker and deploy it somewhere. For people who did not code as project managers, who you're doing things with Excel — this is a pretty big leap — to go from that state to talking about Docker and Kubernetes and all these things. After doing Kaggle, what would be the next step to pick up these skills as well? (40:01)

Ksenia: It also was a question to me — how to jump from analytics to machine learning and machine learning in production. It's a different story. How I overcame this - I learned it at work. This is where you see this enterprise environment, where you see how people work. You also collaborate, and collaborative development is quite different from what you write in a notebook. You have to understand how Git works. You have to use it. You have to merge some branches and so on. It's a different process. You also should understand how to deploy your code and test it, what tests exist and so on. All this practice of development should be also picked up. For this I would recommend some advanced books, like “Clean Code”. It's important to read if you want to improve your coding skills. This is already advanced skills, I would say. You can improve and learn at work — just to observe how people do this, and, in the parallel, you can also read books about this. You're right it is quite a big leap. (41:07)

Alexey: What you're saying is — it's enough to get hired as a data scientist. It's enough to know the theory, do a little bit of Kaggle, know some basics of programming, know how to visualize data to do some data analysis. All these things — this is sufficient to get hired as a data scientist. But then on the job you learn all these production skills. You learn about Docker. You learn about other things. (42:44)

Getting experience while learning

Ksenia: Two more things. What I planned for myself to acquire a job easily after graduating from university — I got into research and I spent three years in a specific domain. It was super beneficial — when you become professional in some specific domain, you know things. You can bring this knowledge to some company. What I learned and my thesis was about fraud detection. I studied this topic for three years and I coded a lot on this topic. I produced some projects on it. It was super beneficial for me to get hired — my next work after university, after the research position, I got hired as a fraud detection data scientist. The theory that I could bring to the company was super beneficial. So I would recommend, if you really want to study — at university, get involved in some research project or get a position in some laboratory. It really helps you to practice. (43:16)

Ksenia: Second, try to be T-shape. It’s an old concept but it's still valid. I would say, you have to know about anything like math, statistics, engineering, analytics and so on. But choose your strengths. Choose your domain. Choose your topic where you can say that “I know things. I know how to do this”. This will definitely help, because you can then put it in your career and use this benefit.

Alexey: That's indeed something that many people overlook — for me personally, it was also the case — having experience in a specific domain. For you, how did it happen that you choose this particular domain of fraud detection? Was it just by coincidence, or you were actively looking to get involved in this? (45:07)

Ksenia: For me, it was an accident. But any topic or domains that you choose, you should be passionate about. Otherwise, you lose motivation pretty soon. In my case, it was when I got a position in university. During the first semester in university, I was active, I wanted to study. One professor noticed me and invited me to work at Deutsche Telekom in his laboratory. He saw that I could deliver and I could help him to get some result on his research. For me, it was surprising because I didn't search for any work at that time. So for me, it was pure coincidence and I just answered “Okay, of course, let's go”. Because I really wanted to get this practical knowledge. And if you can also provide me feedback on my work, it would be so cool collaboration. That's how I get involved in the research. Honestly, I didn't plan it. But now I can definitely tell that it’s worth to do this. If you can — get involved into projects. (45:31)

Alexey: So what you did was — you were active during classes and then a professor noticed you, invited you to the lab. Then, after graduating, you already had some domain experience in fraud detection, and the position you were looking for was also in this area. Was it very helpful for you to have this experience already in this domain? With your conversation in interviews. (46:50)

Ksenia: Yeah, this was purely about fraud detection. “What techniques have I been using so far, what helped a lot?” My thesis was on a really advanced topic. In my thesis, I modified the node2vec technique that has been developed in 2016 for analyzing networks. I applied this technique to identify or analyze money laundering and fraud and anomalies in loan networks. I was working on identifying fraud in monetary transactions. It’s a super-hot and useful topic. Then I applied to the companies that also work in this money laundering, with this fraud and so on. It was a match. (47:23)

Looking for a job

Alexey: It was a very focused approach. You had this domain expertise, and then you focused specifically on companies that are doing this sort of thing. I guess it worked out. Do you remember how many applications you sent? How many interviews did you get? (48:14)

Ksenia: It was a lot. After university, I sent around 50 applications and I got three offers. It is tough. When you try to get the first data science position, don't be desperate. It's hard every time for everyone. Don't be disappointed. (48:35)

Alexey: It means that there were 50 companies. I don't know what your search area is. Was just Berlin or Germany or Europe? But 50 is quite a few. (49:04)

Ksenia: I was quite ambitious. I also applied to big-4 companies and Google — because my thesis was based on the techniques they extensively use, like node2vec. I applied in Zurich. I didn't hope for anything but yeah. (49:21)

Alexey: Big-4 are these big consultancy companies? (49:48)

Ksenia: Exactly. (49:52)

Alexey: We have a question. Did you feel any discrimination as a woman when interviewing? (49:53)

Ksenia: I would say no. Not at all. Every time that I got interviewed, people asked me professional questions on my topic. I don't feel any discrimination at my processes at least. (50:11)

Alexey: Was it mostly in Germany? You said, you also applied to Zurich. (50:34)

Ksenia: Yeah. I didn't get into the interview process in Zurich. But yes. Any interview processes that I got so far, I got in Germany. It was quite professional and it's decent processes. (50:39)

Gap between theory and practice

Alexey: Do you think you saw any difference between the applications that you covered in courses and classes and applications for your applications in the industry? That's a question from Romala. (50:58)

Ksenia: Yeah, it totally corresponds. What I want to emphasize here, I already applied these techniques to practice during my research and this was extremely beneficial. When you just learn techniques and machine learning approaches, and you don't apply it, you don't really understand how it works under the hood. But if you apply it immediately, you'll understand it, you remember it, and you can repeat it. Then, when I went to work, after university, I already knew more or less what I'm going to do, because I already did it. Yes, it's definitely important. It's definitely useful. What I have learned at university is applicable. (51:15)

Alexey: There was no big gap between theory and practice? (52:04)

Ksenia: I would say no, but it also depends on the courses that you choose. I tried to choose practical deep courses, which I know will be definitely useful. Another tip, if you search for work as a data scientist, at the very beginning of your career changing, open and analyze data science job descriptions that you can find on the internet. This way, you will get familiar with techniques with what your employer expects from. This is super important to know in advance what to learn, because you can learn something, for example, MATLAB. It's useful, you can do a lot of stuff in MATLAB, but not many employers require it. Be prepared from the very beginning to not learn stuff that you don't need or not applicable. So, research the job market, and from the very beginning, understand why you do this and what you actually have to do. (52:11)

Learning plan for transitioning

Alexey: Yeah, that's a great tip. Let's say somebody wants to go into data science now. But they don't have enough time or freedom to quit their job and do a masters and spend two-three years doing that. Do you have any recommendations for them? How should they structure the learning plan to be able to still make the transition while working? (53:29)

Ksenia: Yes, I think I can give some tips or some recommendations. First of all, start to be interested in data analysis and start to apply this at your work already. This would be the first step of getting involved in data analysis. Then, when you feel that you can do something or you can derive some knowledge, you can use it for your decision making process. If you don't have time or opportunity to get to university, I would recommend to get involved in some structured courses on data science. I didn't do it myself, but I know some people who did it — nano-degree from Udacity. It lasts for six months. One friend of mine did it and she got some good impression from these courses and she learned a lot. (54:09)

Ksenia: They provide a solid program. The main thing is that they are structured so that you will not get lost during the learning process. Six months is not enough to be a good data scientist, but at least you can follow this path. Then, when you graduate from this, you can get involved in the OpenDataScience course. You have to increase the level of complexity steadily. Don't try to jump into the most deep topic — it's not going to help you. But try to go with small steps and get focused. That's probably how you can approach this path without a university degree.

Alexey: Do you remember which courses your friend took? Was your friend a project manager? (56:19)

Ksenia: She took economics and yeah, she's quite well in Statistics. I think it was the data science nano-degree. (56:28)

Alexey: I don't know much about economics — what kind of tools they use at work, or what they learn. They already know math, and they already know some data analytics. At least basics, like Excel. (56:42)

Ksenia: Yes, she's familiar with statistics as math. It depends on your background. Of course, if you're not familiar with math, you have to figure out how to study it, because it's the basis. (57:06)

Alexey: If you don't know math, you need to find a way to pick it. Do you maybe have any recommendations — which areas? When we say math, it's a lot of things. You can study math for 100 years. Any areas in mathematics that you think are particularly useful? (57:20)

Ksenia: Probability theory for sure, statistics, graph theory. What else... differential equations. That's what I studied. But if you are not familiar with math, studying it by yourself from scratch is super challenging. I would probably go for practical use cases and start to understand it from practice. Because it's better if you do it by yourself, you understand “why?”, “what result?” At least this kind of understanding, this kind of way of thinking — that's what you can develop. Start with theoretical fundamentals of math and then figure out by yourself how to apply it to the problem. Else, it could be extremely difficult. (57:42)

Building on your strengths

Alexey: I think you mentioned that a couple of times already — built on your strengths. If your strength is in project management, it's planning and domain expertise. Use that and try to focus on problems first, and then eventually… What works for me is practice — getting to code as fast as possible, and then trying to code. But I'm an engineer and perhaps this is not the best suggestion for project managers. But use the skills that you already have. Do you know how they can do this? Let's say, a project manager doesn't know mathematics. They don't know how to code. But they have a lot of domain expertise. I think you mentioned learning at work right? (58:43)

Ksenia: Exactly. If you’re a domain expert, and you know marketing. You know that user segmentation is a hot topic and you struggle to find the optimal segmentation to run your marketing campaign efficiently and to attract users that you really want. You know the problem, and you know what you want to achieve. You know your KPIs at least. You already can Google how people usually solve this problem. What are the best practices? Just try to go from the problem — because it's your strength. Try to use it. Go from the use case that you have at work. (59:41)

Alexey: What you did is you really thought carefully about how you structure your learning. This is what you can also do as a project manager — really think about all these bits and pieces. I was taking a course about project management recently and I learned about the concept of “critical path”. So it's important to find this critical path in your learning. And focus on that. (1:00:24)

Ksenia: Yeah. I spent nine years studying in my life. The most important takeaways — learn how to study. You will be studying all your life. Get used to this. If you don’t find an efficient way to structure information, you will be lost. So it's important. (1:01:01)

Last tips and getting involved in projects

Alexey: Do you have any other tips or tricks? (1:01:23)

Ksenia: Get involved in the projects — a voluntary project, study project, open source project. It doesn't matter. That's what I did. I was active, I was involved in some local research, and did some analysis for two years in a row. It's officially published on the internet. Many people have been interested in it and it helped me to progress. I also got a lot of feedback from people around who wasn't very interested in this, and it was really helpful to improve. Another tip so get involved in the machine learning class course, from OpenDataScience. Narrow down your scope. Don't try to be a generalist, like a data scientist and everything. Try to find your strengths. Try to narrow your domain and be active there. When you search for the first job, it's also interesting to look at annual surveys — from O'Reilly, from JetBrains, and other surveys that explore the data science field and ask data scientists, which tool they use, which problems are resolved, and so on. You can find these surveys on the internet. I also did some surveys on this — analyzed and posted also on the internet. Because it's my time when I started to work as a data scientist. I was so interested in “What tools should I learn?”, “What positions should I go” and so on. To get some feelings about this, I joined some projects and I got some research and analysis on this topic. And communities. Join communities. It's important. (1:01:27)

Alexey: Speaking of surveys, there’s also a good one from StackOverflow, even though it's not data science specific. They do a great job. StackOverflow is a very large community, so the amount of information that people share there is amazing. You said one from O'Reilly, and another one? (1:04:14)

Ksenia: JetBrains. (1:04:39)

Alexey: I wasn't aware of this JetBrains one. (1:04:41)

Ksenia: Yeah, I found them last year and it's comprehensive. It's also not in the data science field but its development field. (1:04:44)

Alexey: If you can give me some links. I will put them in the show notes. Do you have any last words? (1:04:56)

Ksenia: Yeah, I wanted this talk to be interesting to people that really want to change their career. Be brave enough, be active. I’m definitely sure that you can tackle it. And I wish you all the best on your career path. (1:05:07)

Alexey: Thanks for joining us today and sharing all your experience, your knowledge and your career transition. Thanks a lot for coming today and sharing it all with us. (1:05:26)

Ksenia: Thank you for inviting me. (1:05:42)

Alexey: And thanks, everyone, for attending and for watching. With that we can conclude and I wish everyone a great weekend. (1:05:44)

Subscribe to our weekly newsletter and join our Slack.
We'll keep you informed about our events, articles, courses, and everything else happening in the Club.

DataTalks.Club. Hosted on GitHub Pages. We use cookies.