About the role
- Develop and enhance the ranking model using the existing SharedBottomModel and multi-task learning framework.
- Engineer user-level behavioral features from ordering and interaction data.
- Integrate GenAI-generated user profile features and embeddings into model training and inference workflows.
- Build and adapt ETL pipelines for feature extraction, transformation, and profile generation.
- Implement projection layers, gating mechanisms, and preprocessing logic for embedding-based features.
- Execute model training, offline validation, and hyperparameter optimization workflows.
- Apply techniques such as inverse propensity scoring to address ranking bias.
- Deploy trained models to SageMaker endpoints and validate integration with the serving platform.
- Support shadow-mode validation, monitoring, model drift awareness, and operational readiness.
- Produce technical documentation and runbooks for deployment, retraining, and support handoff.
- Strong hands-on experience in machine learning engineering and production ML deployment.
- Proficiency in Python, PySpark, and ML pipeline development.
- Experience with Amazon SageMaker for training, tuning, batch processing, and endpoint deployment.
- Familiarity with feature engineering for recommendation or ranking systems.
- Strong understanding of embeddings, model integration patterns, and model-serving considerations.
- Experience working with distributed data pipelines, cloud storage, and production data environments.
- Ability to troubleshoot end-to-end ML workflows across data preparation, training, deployment, and monitoring.
- Experience integrating GenAI or LLM-generated outputs into downstream ML systems.
- Familiarity with feature stores, batch inference pipelines, and retraining orchestration.
- Experience supporting online experimentation or shadow deployments in production systems.
- Exposure to operational monitoring tools such as CloudWatch or model monitoring solutions.
- Prior experience in personalization, marketplace platforms, or consumer recommendation products.