# Tim Kostolansky

> Personal website of Timothy H. Kostolansky, an ML researcher focused on interpretability, AI safety, and understanding how language models work internally. Based in San Francisco.

Tim builds and studies ML systems, with a focus on making model behavior more transparent and reliable. He holds an MEng in Computer Science and dual BS degrees in Physics and Computer Science from MIT.

## Current

- [Workshop Labs](https://workshoplabs.ai): Building AI for a human future (2025-present)
- [Now page](https://tim0120.github.io/now): What Tim is currently working on and recent updates

## Research

- [CoT Red-Handed: Stress Testing Chain-of-Thought Monitoring](https://arxiv.org/abs/2505.23575): Paper on stress-testing chain-of-thought monitoring, accepted to NeurIPS 2025. With LASR Labs.
- [Inverse Constitutional AI](https://dspace.mit.edu/handle/1721.1/156804): Master's thesis on interpretability and steering of language models. MIT, 2024.
- [Entity representations research](https://humancompatible.ai): Investigating entity representations in transformer language models at CHAI (2025)

## Projects

- [Emoji Search](https://tim0120.github.io/projects/emoji-search): An emoji search engine

## Contact

- [GitHub](https://github.com/tim0120)
- [Twitter](https://twitter.com/thkostolansky)
- [LinkedIn](https://linkedin.com/in/thkostolansky)
- [Google Scholar](https://scholar.google.com/citations?user=tzvUuOIAAAAJ&hl=en)

## Optional

- [Resume](https://tim0120.github.io/documents/resume.pdf): Tim's CV/resume
- [Miscellany](https://tim0120.github.io/miscellany): Reading habits, favorite tools, and basketball career
- [Prior Work](https://tim0120.github.io/previous-work): Full list of publications and projects
- [Zotero library](https://www.zotero.org/thkostolansky): Research library
- [Curius](https://curius.app/timothy-kostolansky): Reading list and saved articles