India Governance Watch (IGW)
Public Policy-Intelligence Platform with NLP Ingestion
Public-facing platform that ingests India's Central Government policy PDFs, auto-extracts and classifies individual policy items by ministry/theme/impact via a fuzzy-matching NLP pipeline, and serves them through a searchable dashboard with hand-built D3 analytics - replacing a manual, spreadsheet-based tracking process.
Architecture
Next.js 15.5 + React 19.2 + TypeScript 5 + D3.js/Recharts + FastAPI (Python 3.11) + MongoDB Atlas (Motor, GridFS) + a deterministic TF-IDF/cosine-similarity pipeline (PyMuPDF + RapidFuzz, no LLM calls) + Google OAuth/JWT, on Docker + Cloud Run + Firebase App Hosting.
Key Features
Deterministic, explainable ingestion pipeline (PyMuPDF extraction, RapidFuzz institution matching, custom TF-IDF/cosine similarity) - zero LLM calls, every match traces back to text similarity rather than a model's opinion
Auto-links each new policy item to related earlier ones, forming a policy-evolution timeline per institution
Hand-built D3/Recharts visualisations (treemaps, sunburst, stacked-bar activity charts) over an India map, plus a public dashboard and admin editing tools
47 API endpoints, 109 automated tests run fully offline against mongomock-motor (no live database needed for CI)
Automated fortnightly newsletter pipeline: subscribe/unsubscribe, welcome email, new-issue notifications, admin broadcast
Next.jsFastAPIMongoDBD3.jsNLP