Hey, I'm Tommy! I spent 2.5 years founding and building Clarum — learned a lot about building companies and production AI agents. Realized I'm more drawn to research and contributing to the cutting edge.

Broadly interested in topics that improve our understanding of machine learning systems and their generalization capabilities including:

• Interpretability: mechanistic interpretability, representation analysis, attribution methods
• Open-ended learning: self-play, unsupervised RL
• Mathematical foundations of ML: geometry of learning, optimization techniques, deep learning theory

Email: tommyhe6@gmail.com
LinkedIn: https://linkedin.com/in/tommyhe6
Github: https://github.com/tommyhe6

Writings

• Handcrafted Backdoors for Language Models
• Latent Riemannian Diffusion Models with Mixed Curvature
• Overview of Isogeny Based Cryptography
• Security is not Well-Defined

Notes

Interesting topics I've studied independently.

• Theory of Machine Learning
• Graph Neural Networks and Their Expressive Power
• Introduction to Algebraic Geometry