I build LLM systems and measure whether they actually work. Most recently an Applied Scientist Intern at Amazon (Feb–Jun 2026), where I built LLM pipelines that turned millions of customer feedback records into taxonomies.
- 🔨 Building Decode: turns research papers into educational video, with each stage checking and revising its own output.
- 🔍 Attest: a research agent that checks every claim against the paragraph it cites, and measures what that checking costs.
- 📄 DocuLens AI: grounded Q&A and search over documents, with citations.
- 🧪 Interested in LLM evaluation, taxonomy quality, and efficient inference.