Michael Poli
Co-founder & President at Radical Numerics
Palo Alto, California, United States
Summary
🤩
Rockstar🎓
Top SchoolMichael Poli is a deep learning researcher-engineer and founder with 7 years of experience building numerics-driven ML systems and production-ready model architectures. As Co-founder & President of Radical Numerics and former founding scientist at Liquid AI, he specializes in continuous-depth models, neural differential equations, and scaling large architectures for speed and numerical stability. He co-founded DiffEqML and is a key contributor to torchdyn, a leading PyTorch library for neural differential equations, where his work on Neural SDEs and Galerkin layers highlights a focus on both theory and performant implementation. His background spans industry research roles at Microsoft, NAVER, and startups developing AI for trading and greenhouse automation, showing a knack for translating advanced methods into applied systems. With a PhD in Computer Science from Stanford and interdisciplinary training in control and automation, he blends control-theoretic rigor with modern deep learning practice. An interesting thread through his career is framing ML problems as dynamical systems, enabling novel model families that trade parameter efficiency for richer continuous semantics.
7 years of coding experience
5 years of employment as a software developer
Doctor of Philosophy - PhD Computer Science, Doctor of Philosophy - PhD Computer Science at Stanford UniversityBachelor of Engineering (B.E.) Control Engineering, Bachelor of Engineering (B.E.) Control Engineering at Tongji UniversityChinese (Semester-Long Intensive Language Program), Chinese (Semester-Long Intensive Language Program) at Nanjing UniversityBachelor of Engineering (B.E.) Automation Engineering, Bachelor of Engineering (B.E.) Automation Engineering at Alma Mater Studiorum – Università di BolognaMaster of Science - M.S. Industrial & Systems Engineering, Master of Science - M.S. Industrial & Systems Engineering at Korea Advanced Institute of Science and Technology
English, Italian, Chinese, French, Korean
Github Skills (3)
machine-learning10
pytorch10
data-augmentation9
Programming languages (7)
Github contributions (5)
DiffEqML/torchdyn
Apr 2020 - Dec 2022
A PyTorch library entirely dedicated to neural differential equations, implicit models and related numerical methods
Role in this project:
ML Engineer
ML EngineerContributions:24 releases, 2 reviews, 395 commits in 2 years 8 months
Contributions summary:Michael's commits focus on the development and testing of a Neural SDE model. They are involved in implementing various components of the model, including drift and diffusion functions and the overall integration. They also worked on the use of data control and a Galerkin layer, suggesting a focus on model architecture and performance.
dynamical-systemsdifferentialdedicateddifferential-equationsnumerical-methods
Zymrael/zymrael.github.io
Jun 2021 - Nov 2024
A beautiful, simple, clean, and responsive Jekyll theme for academics
Contributions:1 PR, 26 pushes, 6 branches in 3 years 6 months
responsive-jekyll-themeacademicscleanjekylljekyll-theme
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.