Senior Data Scientist - Recommendation Systems Pod

Overview

Responsibilities

• Design, train, and deploy recommendations/personalization models leveraging deep learning, sequence models (Transformers, GRU), and boosted trees (XGBoost, LightGBM)
• Develop multi-objective ranking that blends engagement, conversion, and merchant value into a single ranking score (value model), using multi-task learning where shared representations help
• Build scalable two-stage retrieval and ranking systems — ANN retrieval (FAISS, ScaNN) over user/product/event embeddings feeding learning-to-rank models (pointwise, pairwise, and listwise objectives)
• Collaborate with infra to productionize real-time feature pipelines (ClickHouse, Kafka, Spark)
• Define serving-time impression and feature logging to eliminate training-serving skew and produce unbiased training data
• Design and run online experiments with rigorous guardrail metrics; correct for position and presentation bias in logged data; apply counterfactual/off-policy evaluation and uplift modeling to attribute lift accurately
• Integrate model outputs with platform APIs for dynamic personalization in search, home feeds, and store pages
• Define best practices for offline evaluation (MAP@K, NDCG) and online experimentation metrics (CTR, CVR, GMV uplift)
• Partner with product analytics and data science to iterate on signal enrichment and cold-start strategies
• Mentor junior data scientists and define best practices

Requirements

• Bachelor's or Master's degree in Computer Science, Machine Learning, or a related technical field
• 4+ years of hands-on ML experience, including 2+ years designing or deploying large-scale recommendation systems
• Track record: Built or maintained systems serving 1M+ users or generating 100M+ personalized predictions daily
• Deep expertise in representation learning, embeddings, attention mechanisms, and multi-task learning
• Demonstrated success integrating multi-stage ranking systems across e-commerce surfaces (search, feeds, product detail pages) with measurable online lift (CVR, GMV)
• Proficient with large-scale data ecosystems: Kafka, Spark, ClickHouse, BigQuery, or equivalent
• Strong command of experimentation rigor: guardrail metrics, position-bias correction, off-policy/counterfactual evaluation, and model monitoring
• Skilled in debugging, optimization, and productionization of ML pipelines in cloud or containerized environments