Project

RADAR

Problem

Practitioners working with domestic abuse survivors in the UK rely on ad-hoc Google searches and phone calls to colleagues to find relevant evidence and guidance. Existing resources (Women's Aid Evidence Hub, SafeLives, Routes to Support) each cover parts of the landscape, but no unified searchable platform exists across both academic and grey literature. A research team at Sussex needed a system that could maintain a living, monthly-refreshed systematic review across hundreds of sources and make the results accessible to frontline workers.

Approach

I built RADAR as two connected systems: an ingestion pipeline and a public-facing portal. The pipeline aggregates content from 487 trusted sources, including academic databases (PubMed, Scopus, Web of Science) via their APIs, government publications via the GOV.UK API, and grey literature from VAWG organisations, local authorities, and children's charities via site-specific adapters. Each source has a dedicated adapter behind a common interface, making it straightforward to add new sources. Incoming documents are deduplicated using a three-tier strategy: DOI match, SHA-256 content hash, then fuzzy title matching (rapidfuzz at 90% similarity). Documents that pass deduplication are screened by the Claude API against the study's inclusion criteria, with a 90% confidence threshold. Items below that threshold are exported for manual review, and every screening decision is audit-trailed with model version, prompt version, and criteria hash for reproducibility.

The portal is a FastAPI application backed by PostgreSQL with pgvector. Practitioners interact through a chat interface where they can ask questions in natural language and receive synthesised answers grounded in the indexed evidence. The pipeline is designed for monthly re-runs with change detection: content hashing ensures only new or modified documents are re-screened, keeping costs and processing time low.

Outcome

RADAR is launching in June 2026 as a public-facing portal. It is the first unified evidence spine for UK domestic abuse services, combining academic and grey literature in a single searchable resource designed for practitioners, commissioners, and researchers.