Aleksey Samosiuk

Alexey is a project manager of the GIS laboratory at MIPT, working on approximate algorithms with a focus on the theory of deep neural networks. The work is used in prototyping a variety of algorithms for the lab's partners: from generative models to classical graph algorithms; from more optimal methods of hyperparameter selection to new approaches to neural network training.

The main focus of the lab is on obtaining prototypes of algorithms that can work in real-world conditions but are not black boxes, even if neural networks are used inside. The lab has had more than twenty such projects. Now Alexey and his colleagues are assembling them into one big project of a new generation of neural networks (beyond the concept of "big model+adapters" towards modularity).