Sam Daulton
Research Scientist at Meta
Truckee, California, United States
Summary
🤩
Rockstar🎓
Top SchoolSam Daulton is a Research Scientist with 11 years of experience, currently a Senior Staff Research Scientist at Meta specializing in Bayesian optimization and adaptive experimentation. He combines deep ML research (PhD-level work in machine learning from Oxford) with hands-on engineering, contributing to prominent open-source projects like Ax, GPyTorch, and BoTorch to enable scalable Bayesian methods and batch evaluation techniques. His work bridges theory and practice—refactoring core GP and multivariate distribution implementations while improving tutorials and user-facing examples to broaden adoption. Based in Truckee, CA, he brings a track record of improving both library internals and educational materials, showing a knack for making complex probabilistic tooling more reusable and approachable.
11 years of coding experience
2 years of employment as a software developer
Doctor of Philosophy - PhD, Machine Learning, Doctor of Philosophy - PhD, Machine Learning at University of OxfordMaster of Science (M.S.), Computational Science and Engineering, Master of Science (M.S.), Computational Science and Engineering at Harvard UniversityBachelor's Degree, Mathematics and Computer Science, Bachelor's Degree, Mathematics and Computer Science at Colgate University
Github Skills (22)
pytorch10
python10
machine-learning10
gaussian-processes10
bayesian10
optimisation10
gpytorch10
optimization10
multi-objective-optimization9
trainings9
linear-algebra9
deep-learning9
modeling9
batch-processing9
jupyter-notebook8
Programming languages (5)
Github contributions (5)
facebook/Ax
Jul 2019 - Dec 2021
Adaptive Experimentation Platform
Role in this project:
Full-stack Developer
Full-stack DeveloperContributions:66 commits, 150 PRs, 60 comments in 2 years 5 months
Contributions summary:Sam primarily focused on improving the MNIST CNN tutorial within the adaptive experimentation platform. Their contributions included updating the tutorial, adding features like specifying the number of workers, and integrating learning rate decay schedulers. They also refactored the code by splitting out non-MNIST-specific logic to make it reusable for other datasets. These changes suggest a focus on enhancing the educational content and improving the flexibility of the platform.
experimentationadaptivesimulationadaptive-experimentation-platformexperimentation-platform
pytorch/botorch
Oct 2018 - Nov 2022
Bayesian optimization in PyTorch
Role in this project:ML Engineer
ML EngineerContributions:3 releases, 35 reviews, 207 commits in 4 years 1 month
Contributions summary:Sam's commits primarily involve implementing and refining Bayesian Optimization techniques using PyTorch for a library focused on Bayesian Optimization in PyTorch. The code changes show a focus on supporting batch evaluation methods and the implementation of the q-NEHVI acquisition function. This includes refactoring code, creating a utility for input transforms, and adding example notebooks using the qNEHVI for a 1-D problem.
pytorchoptimizationmultiobjective-optimizationmachine-learningbayesian-optimization
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.