Research
Current focus
At Hypha, I'm actively researching agent tool design — how the shape of a tool changes what an agent can do reliably.
Published
- AI-Facilitated Software Project Generation from Natural Language Using Curated Code Snippets
Maps natural-language project descriptions to curated, reusable code snippets instead of relying on unconstrained generation — chain-of-thought prompting lifted snippet-mapping accuracy from 30.3% to 43.1%.
In progress
- LongListBench: A Benchmark for Long-List Entity Extraction Under Layout and OCR Noise
An 80-document benchmark for long-list entity extraction from semi-structured PDFs, with ground-truth JSON, OCR transcripts, and a taxonomy of seven failure modes. Targets a known weak spot of LLM extraction: exhaustively recovering every item of a long list from noisy, real-world layouts.
- Agentic browser automation