Never-ending Learning: from PhD in Game Theory to Data Science training
When you look at Sofya Kiselgof, you see a tender delicate woman, charming in her modesty and good manners. At first glance it’s hard to imagine what these slender shoulders can carry. Graduating from one of the most prestigious universities in Russia with a Master of Science Degree in Business Informatics and PhD in Game Theory, having lots of data-based research experience in various fields using Statistical, Machine Learning and Deep Learning approaches, being a senior lecturer and a head of the data analysis group at the university, Sofya decided to take a new challenge. She started a family, moved to Israel and went to look for new opportunities in the field of Data Science. Leaving her academic past behind, but still committed to a field of research, Sofya was eager to find something that will give her both: more expertise in data science and more industry project experience. That’s how she got to Y-DATA Data Science program.
She was already familiar with the Yandex School of data analysis in Russia, recommending it to her strongest students when she was a teacher. Now it was a time to try it herself and become a student again. It was enough to read several posts about Y-DATA program in a Hebrew-speaking community for machine and deep learning on Facebook to understand that the studies can give experience in the practical field, as well as deepen the knowledge of deep learning, big data and machine learning.
Here’s what Sofya says about her decision to enrol to the program:
“I wanted a refreshment of what I had already known, and I wanted to introduce myself to the current state of the field. I had no prior experience in deep learning, I only knew the basics, as in what the neural net is and how it works, but I had never really tried to code something or apply any models at that point. I didn’t have anything to do with the projects related to deep learning either and I wanted to stay up to date on classical machine learning. I looked up some university programs in English, but I wasn’t sure I wanted to go for a full-time master program. It’s strange to go for one having completed a PHD, after all. Y-Data was on my radar due to the fact that I prefer working in the practical field. I didn’t want to go to the academy after the course, and if one doesn’t go for it, one prefers something more practical.”
The application process went well and the exam itself was quite simple to pass, since Sofya took it serious and has already studied much of the materials even before she decided to apply for the program. But soon after the program has started, the pandemic arrived and all of the classes went online.
“I used to manage to do most of my homework en route from Haifa to Tel Aviv but around the time the pandemic hit we started a deep learning course and obviously had to quarantine, everything went online, it was quite a mess. My homework partner from the first semester worked full-time back then, and it was quite taxing for him to go through all the homework. It was a bit easier for me. I have kids to look after, but it’s still less challenging that having to work full-time while studying.”
Sofya is indeed a wonder woman. Nothing can take her off the route. Even when her project partner couldn’t continue collaborating on the research due to some personal reasons, and she, a foreign girl in a new country, interning at a a leading computational biology company was left to her own devices, she still took it positively.
“Loosing a partner on an industry project led to some invaluable practical experience. I still had a mentor and a great team in the customer company, they helped a lot with their advice and direction of analysis related to the biology specifics of proteins that you can find in different bacteria, for instance. They were very helpful, and it’s not like it felt lonesome in there, but I was the sole person responsible for the practical side of it, and one gains a lot of practical experience this way.”
One of Sofya’s project tasks was developing Deep Learning models to predict protein functions in “Evogene”, a company that deals with product development for life-science based industries, incorporating a deep understanding of biology and leveraging Big Data and Artificial Intelligence.
“The industry project part of my experience was pivotal to me, since it gave me an opportunity to feel how it would work out when working on a real project, a real application of my studies. Good thing about my background is that checking dozens of research papers and looking for an already proposed approach is something that I’m used to, even if it is not related to game theory, the methods are the same. I found an approach in the relevant literature that I adopted it to the task, it worked wonders. Deep learning had never been used in this particular domain before, and it turned out it was possible, not to say quite easy, to implement it. If I had to, say, write a better English to Spanish translator, which has a lot of solutions, I wouldn’t be able to make such impact, but it was possible in this case. I also faced a lot of technical challenges, and thus needed to get a better understanding of deep learning, both from the theoretical and the practical points of view. The customer was surprised we actually had practical, working results. They are going to adjust it and use it in their production. They also want to apply it to training in different areas, as well as to other similar tasks. It was my first major real experience in deep learning, so to speak. Hopefully it is also a real asset for international recruitment.”
So now when the program is over and Sofya has some new skills in her pocket, she’s ready to hit the gas and set off on a new work journey. After a Data Science program she feels more certain in client’s needs, limits and expectations and is ready to approach a new task from all angles.
“I always was interested to work with data, process it, find relevant insight, to code, I enjoy this line of work. Y-DATA course just showed me how data science works in the field and proved that this is what I truly want. I’m ok with coding, writing production code, I did it for the models that I developed, but I don’t want my role in data science be all over the place, with sophisticated software engineering as well, because that usually means there is less time for data, for the things that really excite me. My biological project experience confirmed, however, that I can work in different domains, and there were actually only a few moments where my lack of life science degree posed an issue. It’s mostly data work, and one can quickly catch up on the relevant knowledge. Talking about dream job, it would probably be something related to medical and/or biological data analysis. I find many books and articles on the topic particularly interesting. I think there is room for data science application in this field. A dream job would not be constricted by pure research, where the result of your work is just a written article on the topic. I would love to have a job where I get to influence real products and decisions. One might work on the predictive models, when one is able to influence the decisions, or on the products themselves. I do expect that where ever I end up, I am going to be coding as a data scientist.”
After almost 1,5 hour of talking, Sofya doesn’t seem to be tired or bored for a second. It feels like she can continue talking about her field of interest forever, without repeating herself. And for sure, her modesty and delicacy are hiding so many skills and knowledge one can brag about, that she will always look like a person that you will never stop being surprised about. At the same time, she’s so easy-going and simple in explaining things that might seem hard to a non-technical mind, that you immediately want to learn something from her. And she still doesn’t take it for granted.
“I think all my academic and post-academic experience can be attested to the fact that I never said no to any opportunity that was given to me. Someone came up and asked whether I wanted to be the head of data science application group and I agreed immediately, even though I was a second year PhD student at the time. I would also advise not to be afraid of trying. It is always worth it to go into detail and study something extensively, especially if you don’t get it straight away. Find out how it is proven, how it is derived, etc. I regularly see this in my students — you shouldn’t believe your teacher when they say something is right or correct, you should also prove it for yourself. Sometimes your intuition is enough, sometimes you feel like your intuition is telling you the right thing, yet you don’t know the formal proof. It’s totally fine if that’s enough to understand the task at hand. But if one feels something is intuitive because someone smart said it, or convinced one of it, it is not a good thing. Even though data science is an applied discipline, it’s very important to understand the underlying basics.”
