4 часа назад
Тимлид ML (RecSys)
- Удалённо
- Полная занятость
PythonMLA/B testingPyTorchDeep LearningNLPRecommendation SystemsDatingRecSys
Описание
Responsibilities
- Work on a recommendation system and ML portfolio for a dating product with employment under the Labor Code of the Russian Federation.
- Own the end-to-end recommendation system: candidate generation, ranking, re-ranking, cold start, and exploration considering the interests of both parties.
- Be responsible for online metrics, quality and drift monitoring, rollbacks, and degradation analysis.
- Conduct A/B experiments with a product analyst.
- Develop ML projects in Trust & Safety, monetization, retention, and NLP.
- Build a team and processes.
- Interact with the data platform and analytics.
- Design key models and critical pipelines.
- Participate in code development and review.
Requirements
- Experience with production recommendation systems with responsibility for online results.
- Practical experience with deep learning, PyTorch or similar, neural network ranking and retrieval, training on large data, and inference optimization.
- Proficiency in modern NLP models, transformers, embeddings, and fine-tuning.
- Knowledge of classical ML, CatBoost or LightGBM.
- Confident Python and experience deploying models to production.
- Understanding the differences between offline metrics and online effect, biases, feedback, and drift.
- Skills in monitoring, reproducibility, versioning, and data testing.
Nice to have
- Leadership experience.
- Knowledge of online experiments, reciprocal recommendations, and two-sided markets.
- Experience in Trust & Safety, anti-fraud, computer vision, uplift, churn, and bandits.
- Experience with ClickHouse, Airflow, and vector search.
- Publications, reports, or mentorship.