Senior ML Engineer(D.O) - Remote

- Develop, train, and optimize machine learning models using CatBoost, LightGBM, or XGBoost for production environments.
- Design and improve Learning-to-Rank models using pairwise/listwise ranking approaches and ranking evaluation metrics such as NDCG and MAP.
- Analyze and improve probability calibration performance using techniques such as isotonic regression, Platt scaling, Beta calibration, and Venn-Abers predictors.
- Detect, troubleshoot, and resolve train-serve skew and data leakage issues in real-world ML systems.
- Build robust feature engineering pipelines for time-series and event-based datasets with proper temporal validation strategies.
- Collaborate with cross-functional teams to deploy, monitor, and maintain ML models in production environments.
- Work with large-scale datasets using Python, pandas, PyArrow, PostgreSQL, and related ML tooling.
- Contribute to model evaluation, experimentation, and continuous improvement processes.
- Participate in technical discussions, documentation, and weekly review meetings in English.

- 5+ years of professional ML experience, including 2+ years training and shipping gradient boosted models in production (CatBoost, LightGBM, or XGBoost).
- Hands-on experience with Learning-to-Rank (pairwise/listwise objectives, NDCG, MAP) on real datasets — not only classification.
- Solid grasp of probability calibration: ECE, Brier score, reliability diagrams, isotonic / Platt / Beta / Venn-Abers, and the difference between sharpness and calibration.
- Proven track record of diagnosing and fixing train-serve skew in a real system (please be ready to describe a specific incident in the interview).
- Strong time-leakage awareness for time-series / event data: target encoding, rolling features, pedigree-style features, and how to validate them with proper temporal splits.
- Fluent in Python and the standard scientific stack (NumPy, pandas, scikit-learn, PyArrow). Comfortable in a UNIX/Docker environment.
- Working English (written and verbal) sufficient to read/write technical reports and join weekly review calls.

2. Nice to have:
- Experience with sports / pari-mutuel / betting models, or any odds-driven domain.
- Familiarity with handicapping concepts: Quirin speed, pace figures, workout features, class ratings.
- Experience with conformal prediction, Venn-Abers predictors, or distribution-free calibration.
- Bayesian feature engineering experience (used offline as feature generators only, not online).
- Open-source contributions to CatBoost / scikit-learn / calibration libraries.

3. Tech stack:
Python 3.11+, CatBoost, scikit-learn, pandas, PyArrow, FastAPI (serving side, read-only for this role), Docker, Git, uv. Some
PostgreSQL for race data. CI runs on GitHub Actions.

- Receive 100% salary from the onboarding date.
- Participate in company activities: Teambuilding, travel and other activities.
- Work with large and advanced systems, have the opportunity to develop comprehensive technology skills with complex problems, requiring high accuracy.