A content-based recommendation engine that analyses text and metadata using sentence embeddings (MiniLM-L6-v2) to surface genuinely similar titles instead of shallow keyword matches.
- Semantic recommendation engine — transformer embeddings across 12,000+ titles for accurate similarity
- End-to-end NLP pipeline — Pandas and Scikit-learn mapping features into 384-dimensional dense vectors
- Optimised inference — CPU-only PyTorch runtime with GPU dependencies stripped for smaller builds
- Decoupled full-stack app — React and TanStack Router with CI/CD for zero-downtime deployments
React.jsTanStack RouterDjango RESTPyTorchTransformersPandas
