The work
- Develop recommender systems through offline training, online inference and A/B testing.
- Build retrieval and ranking pipelines using collaborative filtering, two-tower models, boosting or neural approaches.
- Create and maintain feature stores, user and item features and behavioural time series.
- Develop Python ranking and recommendation services and optimise latency and throughput.
- Design A/B experiments and offline validation using relevant ranking metrics and statistical significance.
- Monitor feature drift, recommendation quality and business metrics in production.
- Work with customers on personalisation requirements, experiment analysis and solution presentations.
Role-specific skills
- Commercial recommender-system development experience.
- Confident use of Python, pandas, NumPy, scikit-learn and PyTorch or TensorFlow.
- Understanding of collaborative and content-based filtering, ALS, SVD, item2vec and user2vec.
- Experience building two-tower retrieval models and neural ranking.
- Knowledge of LightGBM, XGBoost or CatBoost for ranking.
- Experience with SQL and Spark/PySpark or similar data processing tools.
- Understanding of retrieval, ranking, re-ranking, filtering and experimentation in a recommendation service.
Commercial experience
We consider Junior+, Middle and Senior specialists. Each field requires at least six months of relevant commercial experience. Courses and personal projects can add to it.
Courses and personal projects do not replace commercial experience.
Experience using AI tools in commercial work is a plus, with results checked through tests and review and confidentiality respected.
Tell us about your experience
Describe a recommendation problem, its retrieval/ranking pipeline and evaluation metrics. Explain how you assessed the model offline and in an experiment.
Share your experienceWe connect with specialists in this field for future projects. This page is a role profile, not an open vacancy.