I'm a full-stack software engineer and researcher on the People + AI team at Google DeepMind. I split my time between AI explorables (interactive articles that explain AI research concepts through bespoke data visualizations) and open-source tools and platforms for studying and supporting human-AI interaction. I'm especially interested in data visualization in interactive, educational contexts.
When I'm not coding, I teach classes at my local circus school, where I specialize in handbalancing and contortion.
Best Submission, Workshop on Visualization for AI Explainability (2024)
An introduction to Patchscopes, an inspection framework for explaining the hidden representations of LLMs, with LLMs.
Best Submission, Workshop on Visualization for AI Explainability (2023)
An introduction to grokking and mechanistic interpretability.
An introduction to Sparse Autoencoders, a technique to unpack and understand the hidden representations of LLMs.
An open-source platform for real-time, large-scale behavioral experiments that supports both human participants and LLM-based agents, used to study collective decision-making and human-AI interaction.
A visual, interactive tool for understanding text, image, and tabular ML models. Used to explore model behavior, debug performance issues, and probe prediction reasoning through salience maps, embedding visualization, and counterfactual generation.
arXiv preprint (2025)
CHI Conference on Human Factors in Computing Systems, Extended Abstracts (2025)
arXiv preprint (2024)
Master's Thesis, Massachusetts Institute of Technology (2021)