I’m ready to replace our rule-based “Top Services” carousel with a true, data-driven recommendation engine that can run seamlessly on both our website and our companion iOS/Android app. The goal is simple: every visitor should see services they’re most likely to book, regardless of the device they’re on. Here’s the current landscape: • User data: click-streams, searches, bookings, ratings and basic profile fields are stored in a relational database and mirrored to a JSON feed. • Platforms: React web front-end, Flutter mobile app, RESTful back-end (Node.js) hosted on AWS. What I need from you 1. Model design & training – choose and justify an approach (matrix factorisation, deep learning, hybrid content-collaborative, etc.) that can learn from sparse interaction data yet adapt in near-real time as new bookings come in. 2. Deployment – expose the model behind a lightweight API (Python FastAPI, Flask or similar) with endpoints such as /recommend/{userId} returning ranked service IDs plus confidence scores. 3. Integration support – provide clear docs, sample calls and, where necessary, helper SDK snippets so my team can wire the API into both the React and Flutter clients without blocking on you. 4. Evaluation – an offline notebook illustrating precision/recall or NDCG on a held-out set, and an online A/B framework outline so we can monitor lift after launch. Nice-to-haves include feature engineering in PySpark, use of TensorFlow Recommenders, and deployment via AWS SageMaker, but I’m open to your preferred stack as long as latency stays low and the pipeline is maintainable. If you have shipped a recommendation system for services before, especially across web and mobile, I’d love to see it. Let’s make our users feel like the platform really knows what they need.