Fedor Lebed

Yandex, ML Engineer

Fedor Lebed is a machine-learning engineer at Yandex, where he works on the alignment of Alice AI. His work focuses on measuring and steering model behavior to provide users with helpful, honest, and harmless experiences. On the technical side, he is interested in online reinforcement learning and evolutionary systems as possible next steps toward artificial general intelligence. Before joining Yandex full-time, Fedor was a research intern at the Laboratory of Methods

for Big Data Analysis at the Higher School of Economics. There, he researched applica- tions of world models for accelerating the convergence of continuous-control solvers.

He also worked at CRI Genetics, developing predictive models of ancestral distribution. His bachelor’s and master’s research both focused on optimization: his bachelor’s work investigated optimization approaches for white-box, non-differentiable functions, while his master’s work explored higher-order optimization methods for neural networks. He studied Logical and Combinatorial Methods of Data Analysis in the master’s program at Lomonosov Moscow State University; his studies were interrupted before completion due to the war. Beyond machine learning, Fedor is an engineer at heart, a sketch artist, a gamer, and a motorcycle enthusiast.