jacopotagliabue/recs-at-resonable-scale

Orchestrates end-to-end deep learning recommendation training via Metaflow workflows with GPU parallelization, feature engineering through dbt/Snowflake dataOps, and experiment tracking with Comet ML. Deploys cached predictions to AWS Lambda/DynamoDB for serverless serving, while providing a Streamlit debugging interface for model analysis across item cohorts. Built on NVIDIA Merlin for GPU-accelerated feature engineering and training, enabling a single ML practitioner to manage the full pipeline without DevOps involvement.

241 stars. No commits in the last 6 months.