I'm a Research Scientist at Criteo AI Lab, where I work on machine learning models for recommendation. We have petabytes of data, billions of users, and millions of products to recommend, and we need to scale our algorithms to work under very hard constraints (< 1ms on inference). There are many challenges both from theoretical perspective (counterfactual learning, incrementality, etc.) and practical perspective (incorporating biases, reducing inference, etc.). Feel to check out careers page or send me a message directly, if interested in working with us.
I received a Ph.D. in Computer Science from Skoltech, where I was a part of ADASE group. My thesis is devoted to combinatorial and neural graph embeddings. My research investigates machine learning on graphs, focusing on representation learning, neural network models, isomorphism, and combinatorial optimization. For example, I proposed graph embeddings that are probably isomorphic in the latent space, i.e. there is one-to-one correspondence between graphs and embeddings, -- a property that might be useful for classification problems and data representation. There are several ongoing projects in this area (combinatorial problems with RL, theoretical properties of graph neural networks, high-impact applications of embeddings to drug discovery, medical diagnostics, and recommendation) and if you are interested to collaborate on these subjects, feel free to contact me.