GitHub

The code behind the questions.

A small, curated set of repositories. For everything else, visit the profile.

Open source is part of the argument.

The best way to understand the research is often to inspect the system that made the claim.

Sandbox and harness primitives for coding agents. Zero dependencies, on purpose.

Multimodal agentic RAG with a real evaluation harness. “It felt pretty good” is not a metric.

Noisy social posts to concise, checkable claims—class project to paper to pip install.

Teaching neural networks to rationalize sentiment with human annotations and LIME.