Junior ML/AI Engineer

Huawei Minsk Research Center
ML Research Intern
Project: Developing an agent for solving hard mathematical problems (IMO-level).
- Collected and filtered a large-scale dataset of IMO-level problems from national olympiads and other prestigious
competitions. Applied embedding-based techniques to cluster problems by topic and difficulty.
∙ Researched and experimented with state-of-the-art neural theorem provers and proof assistants (e.g., AlphaProof, GodelProver).
∙ Analyzed failure modes of LLM-based theorem provers to understand which classes of olympiad problems are hardest for them, and outlined techniques to mitigate these weaknesses.