Data Scientist 2

The core responsibilities for the job include the following:

Personalization and Recommendations:

• Build multi-stage recommendation pipelines: candidate retrieval, scoring, and re-ranking across homepage, PDP, cart, and cross-sell.
• Train user and item embeddings that capture behavioral signals and evolving preferences.
• Measure every change through rigorous online A/B experiments.

Search and Ranking:

• Build and improve ranking models that make search results more relevant and drive more conversions.
• Own query understanding, intent detection, and semantic search.
• Design neural re-ranking using dense retrieval and cross-encoders.

Demand Forecasting and Assortment Planning:

• Build models that predict demand at the SKU, category, and regional levels across short and long horizons.
• Help the business decide which products to stock, in what quantities, and for which segments.
• Build assortment intelligence systems that recommend new products to onboard based on demand gaps and market signals.

Inventory Planning and Optimization:

• Build models to predict stockouts, overstock risk, and optimal reorder points.
• Work with operations and category teams to reduce holding costs while maintaining availability.
• Design end-to-end inventory optimization systems factoring in lead times, demand variability, and supplier constraints.

GenAI, Long-Term Memory and Next Best Action:

• Build LLM-powered features: beauty advisors, product Q& A, and personalized content using RAG pipelines grounded in Purplle's catalog and user data.
• Design long-term memory systems that remember skin type, preferences, and past interactions to power next-best-action.
• Build reflection layers so the AI learns from past user journeys and continuously improves its recommendations.

MLOps and Model Governance:

• Build and maintain feature stores, model registries, and drift-detection pipelines.
• Implement automated retraining and shadow-mode evaluation.
• Every model must be observable, versioned, and rollback-capable in under 5 minutes.

Experimentation and Causal Measurement:

• Run A/B tests and online experiments to measure the real business impact of model changes.
• Design experiment frameworks with power analysis and metric guardrails.
• Apply causal methods, uplift modeling, and synthetic controls to go beyond naive A/B lift.

Requirements:

• DS2 3-6 years; end-to-end ownership of production ML systems and experimentation platforms.

Tech Stack:

• Languages: Python (primary), SQL.
• ML Frameworks: TensorFlow, PyTorch, scikit-learn, XGBoost, LightGBM.
• MLOps: Vertex AI (GCP), MLflow, Kubeflow Pipelines, Databricks.
• Feature Engineering: Feast, BigQuery ML, and dbt.
• Experimentation: Internal A/B platform, statsmodels.
• Data Processing: Apache Spark, BigQuery, Airflow.
• Infrastructure: Kubernetes, Docker, Terraform.
• AI Productivity: Cursor, Claude, GitHub Copilot.

Skills

BigQuery, Databricks, Kubeflow, LightGBM, MLflow, PyTorch, Python, Spark, TensorFlow, XGBoost, scikit-learn