I’m looking for a data-savvy partner who can build a full-stack AI/ML engine that delivers custom recommendations for a retail environment. Unlike the usual off-the-shelf product, content, or service suggesters, this system must analyse our historical sales, inventory movement and operational data to surface insights that directly improve production efficiency—think smarter stock replenishment, leaner order batching and faster in-store fulfilment rather than just “people who bought X also bought Y.” You’ll have clean CSV exports plus real-time feeds ready for training, and I’m open to modern Python tooling—TensorFlow, PyTorch, scikit-learn, even Neo4j or Spark if graph or distributed processing helps. The final model should run behind a REST or GraphQL API so our existing POS and planning dashboards can query it in milliseconds. Key deliverables: • Data-prep and feature engineering notebooks (Jupyter preferred). • A reproducible training pipeline with hyper-parameter search. • The trained model saved in a portable format (ONNX or Pickle). • An inference microservice (Docker container) exposing the recommendation endpoint. • A brief read-me and demo script showing the engine improving a sample production-flow scenario. Acceptance criteria: when we run the demo against last quarter’s data, the engine should suggest actions that reduce forecasted stock-outs and over-production by at least 10 % versus our current rule-based baseline. If you have prior experience melding retail ops data with machine learning and can iterate quickly, let’s talk timelines and milestones.