Alkera AI, Inc. (YC-S26)
Building reinforcement-learning infrastructure and evaluation systems for an LLM data agent.
Michael Yao
Computer Science & Mathematics · Harvey Mudd College
I work at the intersection of computational physics, machine-learning interpretability, and molecular machine learning. My research pairs theory and simulation with careful research software to make complex scientific and learned systems easier to measure, test, and understand.
Research & Experience
Ersilia Open Source Initiative
Developing Chem-ICL, an in-context learning pipeline for molecular-property prediction.
Zhuang Group — Harvey Mudd College
Statistical Physics of Deep Neural Networks
Using mean-field theory to study signal propagation, criticality, and trainable depth in random neural networks.
Read more
Water × Unsupervised ML
Using unsupervised learning to identify latent structural populations in molecular-dynamics simulations of supercooled water.
Read moreAlgoverse
Tracing how reinforcement learning reshapes language-model representations with aligned sparse autoencoders.
Selected Software
SPEQTRO
An autonomous agent that combines NMR, IR, and mass-spectrometry models into ranked molecular structures with traceable evidence.
Chem-ICL
An in-context TabPFN pipeline that combines Ersilia model representations for molecular-property prediction without task-specific training.
ms-pred × PyG
A PyTorch Geometric port of ICEBERG’s MS/MS fragmentation pipeline across eight graph-neural-network families.
Education
Harvey Mudd College
Relevant coursework: Statistical Mechanics, Probability, Statistics, Differential Equations, Graph Theory
Shanghai Qibao Dwight High School
Awards:
- UK Chemistry Olympiad: Round 1 Gold Medalist (2023, 2024), Silver Medalist (2022)
- Canadian Chemistry Olympiad: Round 2 Gold Medalist (2024), Round 1 Gold Medalist (2022)
