Email: t.chaipat AT columbia.edu
I am a senior at Columbia University, studying astrophysics (and mathematical probability). Recently, I have been thinking a lot about interpretability, representation learning, and agentic systems for science.
I have used tools from statistics and physics to build robust, trustworthy ML systems for scientists. My work with the Learning the Universe collaboration focuses on developing a robust parameter-sampling algorithm and a degeneracy detection framework for simulation-based inference.
Aside from research, I have a Substack where I write about AI, philosophy, and the anthropology of progress.
Agentic Systems for Calibrating Cosmological Simulations
Chaipat Tirapongprasert and Matthew Ho
9th Annual Workshop on Machine Learning and the Physical Sciences (ML4PS)
[ Paper]
Don't Cut Corners: How Training Outside the Prior Makes Simulation-Based Inference More Robust
Chaipat Tirapongprasert and Matthew Ho
PRX Intelligence (In press)
[ Paper]
Learning at the Edge: Tailed-Uniform Sampling for Robust Simulation-Based Inference
Chaipat Tirapongprasert and Matthew Ho
2026 Conference on Physics and AI (PAI26)
[ Paper]
DegenDetector: Symbolic Recovery of Parameter Degeneracies in Bayesian Posteriors
Chaipat Tirapongprasert and Matthew Ho
2026 Conference on Physics and AI (PAI26)
[ Paper]
Let Them Talk: A Multi-Level Analysis of Communication Between LLM Agents [ Poster ]
Agentic Systems for Calibrating Cosmological Simulations [ Abstract | Slides | Poster ]
Agentic AI: From Zero to Infinity [ Slides ]
Build Your Own AI Tutors and Research Agents
Detecting Degeneracies for Simulation-based Inference [ Slides | Poster ]
Tailed-Uniform Simulations for Simulation-Based Inference [ Slides | Poster ]