I have a production Node.js platform that now needs a robust AI/ML layer written in Python and exposed cleanly to the existing services. Your mandate is twofold: first, build or refine the core models themselves; second, integrate those models so they can be consumed seamlessly by our live Node codebase. All modelling work should be done in TensorFlow, taking advantage of its ecosystem for training, serving, and monitoring. You will own model development end-to-end—data ingestion helpers, feature engineering, training, evaluation, and versioning—then wrap the final model in a secure API (REST or gRPC) or a lightweight microservice so our Node backend can call it with minimal latency. Code quality, reproducibility, and clear documentation are non-negotiable. Everything must run inside containers and plug straight into our CI/CD pipeline. Deliverables 1. Reproducible TensorFlow training script and notebooks 2. Trained model artefacts plus automated evaluation report 3. Containerised inference service exposing documented endpoints 4. Integration pull request into the Node repository 5. README detailing setup, deployment, and usage A short kickoff call will clarify data access and architectural nuances; after that I expect iterative checkpoints until the integration is merged and live.