Machine Learning Engineer
RAG · NLP · LLM integrations · ETL for ML
Moscow · tsygantsov.ru · LinkedIn · Habr
I ship models and the data around them: retrieval, evaluation, Python services, SQL. Currently building Voicaj — speech and text become structured records (tasks, calendar, health, finance) via YandexGPT, LangChain, and intent-style routing. Previously: corporate RAG (~60% faster document search) and data-quality tooling (~40% fewer bad loads).
M.S. Machine Learning @ Finance University (2026–2028).
| Project | Description | Link |
|---|---|---|
| Voicaj | Voice → structured life data. Next.js, Node/Express, PostgreSQL, Redis, YandexGPT + SpeechKit, Docker, GitHub Actions. | voicaj.ru |
| Heart Risk ML | Binary heart-attack risk from clinical tabular data. CatBoost + FastAPI. | repo |
| Text-to-Image Search | ResNet50 + TF-IDF + MLP · поиск фото по текстовому описанию. | repo |
| TeleDom Retention | TeleDom Retention Model — прогноз оттока абонентов (ROC-AUC 0.976). | repo |
| Retail Retention ML | ML for e-commerce retention: purchase-activity decline, segmentation, personalized offers. | repo |
| Marketplace Sales ML | ML for e-commerce retention: purchase-activity decline, segmentation, personalized offers. | repo |
| HR Analytics ML | HR analytics ML project (workforce / turnover modeling). | repo |
Writing: RAG without a trust policy — rules vs materials, retrieval gates for product assistants.
Python · PyTorch · scikit-learn · LangChain · SQL · PostgreSQL · FastAPI · TypeScript · Next.js · Docker · Airflow · MLflow
tsygantsov.ru · linkedin.com/in/tsygantsov · habr.com/users/luvgreyair · kaggle.com/atsygantsov
luvgreyair@vk.com



