I build models and everything around them: data pipelines, services, CI and monitoring. Lately mostly RAG, LLM services and their evaluation.
MEPhI master's coursework:
- pd-model-mlops - Credit card default model with DVC, MLflow, FastAPI in Docker and drift monitoring
- ecology-helper - Climate RAG chatbot with E5 embeddings, Qwen2.5-0.5B and Streamlit
- sber-avtopodpiska - Conversion prediction on Google Analytics sessions using CatBoost and LightGBM
- ml-kursovaya - Anti-influenza compound activity prediction using RDKit descriptors
Personal:
- camoufox-profiles - Browser profile manager on top of Camoufox with pinned fingerprints and proxy support, native on Linux arm64
- TodoList - C++ CLI task manager with categories, priorities, undo and JSON storage
LLM / NLP: Transformers, PEFT (LoRA), LangGraph, vLLM, FAISS, BM25, cross-encoder reranking, sentence-transformers, Rasa
Data / MLOps: Dagster, DVC, CatBoost, LightGBM