Zach does LLM interpretability and software
I'm working on independent mech interp research these days! Open to consulting software work.
| Quick Links | |
|---|---|
| Resume | /resume |
| hey@zachmaas.com | |
| /in/zach-maas | |
| GitHub | @zmaas |
Current Highlights
| Post |
|---|
| Research Notes: Sequence to Read Models: Notes on training biological transformer models from my graduate work |
| Project | Stack |
|---|---|
| Graphical Activation Probes: Finding structure in sparse LLM representations. Trying to scale up efficiently. |
Python, PyTorch, HuggingFace, SKLearn |
| Memeory: Semantic search on all the pictures you have saved on your phone. v1 built in 48hr. Figuring out local image embeddings. |
Expo, React Native, Typescript, MLKit |
| This Website: static site + progressive enhancement |
Eleventy, Typescript, Cloudflare Workers |