Step by step: from Skoltech Master's to Director at one of the largest banks in Russia
Alexei Riabykh, 2019 MSc in Data Science | Managing Director of Modeling and Data Analysis, VTB
Meet Alexei Riabykh, Managing Director of Modeling and Data Analysis at VTB. He came to Skoltech with ambitions in the fintech sector and here he got the chance to look at his field of study from different angles that allowed him to build a solid base for further development, which led him to his current position.
Enjoy his story!
why I chose Skoltech
First of all, I was really impressed by the professors and classes in my field of study. By that time, I had graduated with a bachelor's degree from the Moscow Institute of Physics and Technology (MIPT) and I wanted to continue studying. But I didn't want to do so just to get a degree, as is often the case with master's programs. I wished to acquire new knowledge and become better at what I do.

Next, I think Skoltech has something that education in Russia lacks. That is, the chance to look at your field of study from different angles: scientific, technical, economic, and innovative points of view. This fact offers you a whole new spectrum of collaboration possibilities. That was really important to me since I already had some ambitions in the fintech sector.

And finally, networking. Of course, you meet some exceptionally talented people at the MIPT. But they're all somewhat aligned in their interests and worldviews. But Skoltech gets students from all walks of life and professions — from biology to engineering. Just meeting and speaking to these people is the perfect way to broaden your horizons.

All in all, I knew what I was doing when I chose to go to Skoltech.

Data scientists are going to be among the most demanded specialists in the hi-tech market. The purpose of our program is to meet this demand and to equip the most talented young scientists with high-level knowledge and experience in machine learning, deep learning, computer vision, industrial data analytics, natural language processing, mathematical modelling and other important areas of modern data science.
Learn more >>
current job
I am a Managing Director and a Team Leader at VTB ML Center. My team mainly deals with tasks related to time series and natural language processing. These may include scanning news feeds and social media for material indicating that the credit rating of a bank's counterparty decreased. For example, some regional newspaper or a Telegram channel says that a factory failed to pay the worker's wages, but the factory belongs to a partner of a firm that the bank regularly lends to. Our models detect this and signal the bank of possible risks factors that may cause late payments or threaten the bank's reserves. So, our machine learning models deal with identifying risk factors, summarizing the news, combining several news items into future scenarios, and forecasting the impact on economic indices. Another example: a retail investor had their asset yield decrease or increase. Our algorithms determine why did this happen by identifying the links between historical news and changes in asset costs and rendering the information in a simple, readable way.

These are just two recent cases, but we have had a huge number of time-series and text processing tasks.

practical application of what I learned at Skoltech
I was a part of the ADASE research group, and Evgeny Burnaev was my advisor. I worked with multivariate forecasting, time series clustering, anomaly detection, fault detection, and discrete optimization. My findings were used in commercial software for cash management optimization (forecasts and recommendations for banks: distribution of cash among ATMs for lowering funding and cash-in-transit costs, improving client service, and satisfying demand). Skoltech professors have a lot to offer in terms of data science and other useful subjects. So as far as hard skills are concerned, I use them daily. This is the foundation for my job.

Skoltech also teaches you to see new opportunities at the intersection of areas. Our team swears by the Data Fusion approach (it presupposes achieving the greatest business effect by combining data and using machine learning algorithms in different areas). There is also a conference of the same name ( by VTB and the Skolkovo Foundation. So our whole team lives by Skoltech principles.
my typical day and responsibilities
My job is remote for now, so I wake up, do my exercises, eat breakfast, open my laptop, check my work emails, estimate the number of meetings that I have today, and make a plan for the day. During the workday, I have different tasks to solve, team members to call about our active projects, and clients to discuss our products with. I always allocate some time for research, writing studies, and coding (I try to balance my soft and hard skills). At the end of the day, I summarize and compare what I managed to accomplish to what I planned to do.

Managing a team is somewhat more difficult when you all work remotely. Daily calls via Zoom and Google Meet, strict KPIs for the team and yourself — it's all fine and dandy, but I feel like I miss being able to share expertise with my colleagues in a semi-formal setting, like at lunch or a tea break. We're trying to fill this void by holding teambuilding exercises outdoors — for example, last time we went to the lakes near Moscow. Can't say we're doing it regularly, but we try.

My direct responsibility is managing the team and the machine learning part of our projects: from identifying the need for machine learning solutions to building something that will satisfy this need.

my career journey and advice to current students
My career began at the Institute for Information Transmission Problems of the Russian Academy of Sciences, which is a joint department of MIPT where students can work in scientific teams. I worked on 2D and 3D computer vision problems and studied attention models of deep neural networks.

After that, I did a lot of research and project work at the CDS Office of Sberbank — their flagship data science division, which is responsible for the digital transformation of Sberbank, among other things. We developed recommendation systems for bank products based on geographical, transactional, and click-stream data. We also held competitions and hackathons and created internal libraries and services.

Next was Gazprombank. At the time, the company's management focused on the retail segment, which produced a whole cascade of modeling tasks. What would be the optimal placement for offices and ATMs in some random town? What factors would ensure good customer flow? Would the placement be optimal in terms of infrastructure and operating costs? So we developed a geoplatform that would collect and model all geodata that the bank acquired.

And now I work at VTB. The tasks of our team (in the broadest sense) are extremely numerous — we deal with all non-standard modeling tasks and foster mutually beneficial partnerships with Big Data providers in Russia. In just two years, we have created the ML Center, which is now capable of solving any data science problems that the modern fintech sector may face.

There were no sudden ups and downs in my career. Every step I took would expand my previous experience and build upon it. And I think that is really important. I believe that it is crucial to constantly find new ways to better yourself and grow at your own pace. And it's especially great if your colleagues and people around share this idea.

My advice to students: Skoltech is a great social elevator. It offers great opportunities, just be sure not to miss them! And, of course, the most important thing is to have fun in the process:)
and now about Skoltech
We are Skoltech – a new international English-speaking STEM university that was founded by the group of world-renowned scientists in 2011 in Moscow, Russia. In just 8 years, we united dozens of researchers and globally renowned professors, built a stunning campus, set up world-class labs and made it to the top 100 young universities in the Nature Index. Read more >>
Discover Skoltech in less than 3 minutes
more stories? sure: