Website Recommendation Engine Build

I need an AI-driven recommendation system that studies on-site user behavior and serves up relevant products and content in real time on my website. The core job is to design, train, and deploy a model that can interpret click-stream, dwell-time, and purchase history data, then return ranked suggestions through an easy-to-call API or directly inside the site’s codebase.

Here is what success looks like to me:
• A behavior-based algorithm (collaborative, content-based, or hybrid—you can advise on the best fit) developed in Python using libraries such as scikit-learn, TensorFlow, PyTorch, or a framework you prefer.
• A lightweight service—REST or GraphQL—that exposes “getRecommendations(user_id)” so my front-end team can drop it straight into our existing stack.
• Clear documentation covering data schema, model training pipeline, and deployment steps so we can retrain as new traffic patterns emerge.
• An evaluation report comparing at least two model approaches and the metrics (precision, recall, MAP, or similar) that justify the final choice.
• Smooth website integration: once deployed on our server or cloud instance, the engine should respond in under 200 ms for a typical request.

I will provide anonymized behavioral logs and can arrange secure database access. If you have prior experience building recommender systems for e-commerce or content platforms, I’d love to see a brief example or demo link when you respond.