Projects
Document Intelligence Pipeline
Multi-format ingestion → extraction → normalization → graph construction.
- Batch + async
- Confidence scoring
- Neo4j-ready graph output
Graph + Vector Retrieval Stack
Hybrid retrieval patterns for deep, connected domain knowledge.
- Qdrant + Neo4j
- Structured queries
- Production-minded design
Developer Workbench / Prompt Ops
Local-first tooling for prompt iteration, eval, diffing, and libraries.
- TypeScript shell
- Registry patterns
- Low tech-debt