Du Phan is a Senior Software Engineer with 11 years of experience building modular, high-performance platforms for probabilistic inference and uncertainty-aware ML systems. Based in Providence, he drives production and research infrastructure at Google—authoring Coix and leading efforts to compose LLM agents and parameter-efficient fine-tuning—while remaining an active open-source core contributor to NumPyro, Pyro, JAX, and TensorFlow Probability. His background (PhD in mathematics) and hands-on work span distributions, samplers, and GP/uncertainty layers, reflecting deep expertise in both theoretical statistics and practical systems engineering. Colleagues know him for improving numerical stability and shape handling in foundational libraries and for translating advanced Bayesian workflows into reproducible tooling for research and product teams. An avid problem-solver who describes himself as “a math student, a data novice, a zen mind,” he blends rigorous math with pragmatic code to make probabilistic methods usable at scale.
11 years of coding experience
4 years of employment as a software developer
Full Stack Web Development Certification, Computer Software Engineering, Full Stack Web Development Certification, Computer Software Engineering at freeCodeCamp
Mathematics, Mathematics at VNU-HCM High School for the Gifted
Doctor of Philosophy - PhD, Mathematics, Doctor of Philosophy - PhD, Mathematics at Pohang University of Science and Technology
Bachelor of Science - BS, Mathematics and Computer Science, Bachelor of Science - BS, Mathematics and Computer Science at VNUHCM - University of Science
Probabilistic programming with NumPy powered by JAX for autograd and JIT compilation to GPU/TPU/CPU.
Role in this project:
Data Scientist
Contributions:30 releases, 1693 reviews, 767 commits in 3 years 11 months
Contributions summary:Du's commits focused on enhancing the `numpyro` library, specifically within the domain of probabilistic programming and Bayesian inference. Their work involved integrating and utilizing upstream sampler versions, including comments and code modifications to the distributions and transforms modules. These contributions involved improvements to specific distribution implementations, suggesting expertise in applied statistics, Bayesian inference, and the use of probabilistic programming tools and techniques.
Deep universal probabilistic programming with Python and PyTorch
Role in this project:
Back-end Developer
Contributions:1 release, 255 reviews, 254 commits in 5 years 2 months
Contributions summary:Du primarily contributed to the development of the pyro-ppl/pyro library, focusing on core features such as adding and improving distributions, and implementing Gaussian Process (GP) related functionalities. Their work includes adding a Binomial distribution, integrating a sparse multivariate normal distribution, and implementing sparse variational GP regression models, along with a tutorial. Additionally, the user addressed a bug related to the sparse tests and the initialization of the HMC (Hamiltonian Monte Carlo) algorithm in the library.
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.