AI Recommendation System Capstone

I’m working on my graduate-level capstone and have chosen to build a full-fledged recommendation system using a publicly available dataset (e.g., MovieLens, Amazon Reviews, Goodreads—whichever fits best once we outline the approach). The end goal is to demonstrate solid grasp of modern recommender techniques, from data preprocessing through model training, evaluation, and an easy-to-showcase demo. Scope of work The project has three pillars: 1. Data pipeline – ingest, clean, and transform the chosen public dataset so it’s ready for modelling while remaining fully reproducible (Python, pandas, SQL or Spark if scale requires). 2. Modelling – experiment with at least two algorithms that illustrate contrasting paradigms, for example matrix factorization (Surprise, implicit, LightFM) versus a neural approach (TensorFlow/PyTorch). Hyperparameter tuning and offline metrics such as precision@k, recall@k, MAP, or NDCG should be clearly reported. 3. Demo & report – a lightweight web or notebook-based interface that lets a reviewer type or click an item and instantly see the top-N recommendations, plus a concise technical report summarising methodology, results, and next steps. Acceptance criteria • Code runs end-to-end on my machine with a single command or notebook run. • Evaluation metrics and comparison table included. • Read-me explains how to reproduce results and launch the demo. • Final write-up (≈10 pages) meets academic standards and is citation-ready. Timeline is flexible enough for thoughtful experimentation yet tight enough to keep momentum; we can break delivery into milestones (data prep, initial model, final model & demo, report). If you are comfortable with recommender libraries, Python, and clear scientific communication, let’s discuss details and lock the dataset choice so we can start right away.

Project ID: 40483996