Machine Learning Engineer, Marketplace
Responsibilities
-
Analyse user behaviour data (purchase history, browsing patterns, game genre preferences, session signals) to identify high-value personalisation features
-
Design, train, and iterate on recommendation models — from collaborative filtering and matrix factorisation to sequence-based and embedding-based approaches
-
Build and maintain end-to-end training and serving pipelines in collaboration with data and backend engineers
-
Define and track evaluation metrics — offline (precision@k, NDCG, coverage) and online (CTR, conversion, revenue per session) — tied directly to business KPIs
-
Run rigorous A/B tests to benchmark new approaches against the current internal baseline
-
Own monitoring and observability of deployed models: data drift, prediction distribution shifts, latency, degradation
-
Contribute reusable user and item features to our feature store
Requirements
-
Hands-on experience designing and shipping recommender systems — collaborative filtering, content-based, hybrid, or sequence-based. You've gone beyond tutorials and built things that shipped and improved real metrics.
-
End-to-end ML ownership — you've taken models from raw data through feature engineering, training, evaluation, API wrapping, deployment, and production monitoring. You don't hand off at the notebook stage.
-
Strong Python and MLOps fluency — extensive Python for model development, plus experience with MLOps tooling (MLflow or similar) for experiment tracking, model versioning, and lifecycle management.
Nice to have
-
Experience with real-time or streaming inference (Kafka, Flink) for session-based recommendations
-
Familiarity with Databricks and/or Apache Spark for large-scale data processing
-
Production experience with feature stores (Databricks Feature Store, Hopsworks, Feast, or similar)
-
Knowledge of two-tower / embedding-based retrieval at scale
-
Familiarity with bandit algorithms or reinforcement learning for online recommendation optimisation
-
Strong business communication skills — you can translate model results and experimental findings into clear, actionable language for product and commercial stakeholders.
