I recently co-founded an AI Safety Research Infrastructure Organisation (501c3, non-profit) which aims to accelerate progress in mechanistic interpretability, in the service of mitigating risks from future AI systems.

Our projects include:

1. Neuronpedia: An open platform for interpretability research. This site hosts sparse autoencoders trained by independent researchers as well as OpenAI and DeepMind.

2. SAE Lens: A library for training and analysing sparse autoencoders.

Prior to my current role, I was an independently funded AI Alignment Research Engineer focussing on mechanistic interpretability in reinforcement learning. I was previously a maintainer of TransformerLens, a popular open source package for mechanistic interpretability of transformers.

Prior to working in AI Alignment, I studied computational biology, and worked for 2 years as a data scientist in a proteomics startup.