Yidi Qi
Yidi Qi is a postdoctoral fellow in applied mathematics at Harvard University, working in the Geometric Machine Learning group and supported by the DARPA expMath program.
His research focuses on developing AI methods for mathematics and physics. He currently works on automated curriculum learning for mathematical discovery, and on building autoformalization tools to generate machine-verified proofs. He also uses neural networks and evolutionary algorithms to study Calabi-Yau geometry.
He completed his PhD in theoretical physics with Fabian Ruehle at Northeastern University and IAIFI. He earned his master's degree at Stony Brook University, advised by Michael R. Douglas.
His life goal is to develop ideas and projects for the future artificial superintelligence (ASI) to learn from before it outsmarts us.
News
Research
My research follows three complementary pillars covering the full spectrum of mathematical inquiry:
AI for Approximation
Solving complex PDE systems and perform calculations that are intractable with traditional methods, such as Calabi-Yau and G2 metrics.
AI for Discovery
Exploring novel patterns, structures, and conjectures across vast physical and mathematical landscapes.
AI for Rigor
Bridging the gap between numerical approximation and formal certainty, creating pathways for computer-assisted and formally verified proofs.
Key Projects
Choir
An open protocol for distributed multi-agent autoformalization.
MLGeometry
An open-source Python package for computing Calabi-Yau metrics.
Publications
Authors of mathematics papers are listed in alphabetical order, following the field's convention
Loading publications...