Project case study · 2026
Sentinel Geopolitical RAG Platform
A full-stack intelligence system that turns continuously ingested reporting into grounded, source-cited analysis.
View project on GitHubProblem
Analysts need a way to search a changing news corpus while retaining source traceability.
Approach
Sentinel ingests news and research sources asynchronously, normalizes and semantically chunks articles, embeds them, ranks retrieval results, and generates answers that retain citations.
Architecture
scheduled ingestion → enrichment and chunking → ChromaDB + PostgreSQL → retrieval and source ranking → answer generation → React dashboard
Implementation
FastAPI services coordinate ingestion and retrieval, Redis provides caching, PostgreSQL keeps persistent metadata, and a React interface exposes the workflow. Docker packages the system for deployment.
Evidence
The implementation demonstrates the complete flow from scheduled ingestion to retrieval and grounded answers with source citations. This is system-level evidence; it does not imply a numerical accuracy result.
Limitations
Retrieval quality depends on source coverage, chunking, and the quality of upstream publications.