I studied Computer Science and Classics at Dartmouth, and I work as a healthcare data analyst. In my own time I make software for reading and learning.
Read real Latin with help on every word: the meaning, the parse, and a question that teaches. I built it alone, from the morphology model to the reader.
A local-first book reader. Mark each highlight with one keystroke for what it is, then turn your marks into Claude-drafted flashcards that sync to Anki.
Graph-RAG code search across 300K+ lines of code: Neo4j code graphs, embedding retrieval and LLM synthesis. I architected it and designed the evaluation benchmark (team of 5).
An LLM adapter that combines LoRA with dynamic prompt embeddings via cross-attention (team of 2). Beat Alpaca-LoRA on PIQA and WinoGrande, and trains in about 5.5 hours on two RTX 4090s.
Write-up coming soon.
I came to computing through Classics, and most of what I build comes out of that: tools for reading difficult texts and remembering what you read.
Product, engineering, model pipeline, deployment and user acquisition, end to end.
Cost modeling and reporting for pharmacy-benefit clients.
B.A., Computer Science modified with Classical Languages. Henry C. Beck Jr. 1938 Scholar.
Kemeny Computing Prize (System Design) · Atherton 1883 Greek Prize · Magnuson Center for Entrepreneurship grant for antiQ.ai · Academic Excellence Citation, Multi-Modalities Generative AI
Financial modeling, valuation, accounting and corporate finance.