I work on getting more and higher quality tokens for LLMs. I used to work on programming tools for data exploration and processing.
🎓 PhD in databases at UC Berkeley'20. BA in Computer Science at Harvard College'14.
🏗️ I currently work at Anthropic. From 2021 to 2022, I was co-founder and CTO of Linea Labs. I had briefly been at Microsoft Research, Clever, Twitter, and Microsoft.
📢 I have given talks at StrangeLoop, JupyterCon, ForwardJS, Papers We Love, and industrial/academic research groups.
Scalable interactive visualization DIEL developed two ideas: (1) declarative hybrid execution between client DB (sql.js) and server DB 📄 Transparent Scaling for Interactive Visualization, (2) reactive programming using SQL queries while resolving asynchrony through DB transactions.📄 Interactive Visualization Beyond the Here and Now (more on the consistency piece: 📄 A DeVIL-ish Approach to Inconsistency in Interactive Visualizations). 🤝 Work done with Remco Chang, Joe Hellerstein, Arvind Satyanarayan, and Eugene Wu. ⭐ OSS code.
Adding interactions to literate programming B2 is bolt on interactive visualization tool to dataframes in notebooks; supporting interactions and coding 📄 B2: Bridging Code and Interactive Visualization in Computational Notebooks. 🤝 Work done with Joe Hellerstein and Arvind Satyanarayan. ⭐ OSS code.
Developer UX for data I 💙 designing abstractions and studying the ergonomics of APIs. Is a Dataframe Just a Table. It resonated with the Hacker News community and hit the front-page.
Python program analysis Lineapy is a Python program slicer designed to clean up messy notebooks into data pipelines. 🤝 I worked with Saul and the rest of the Linea team. ⭐ OSS snapshot of the code before my departure.