Sebastian Ament
Staff Research Scientist at Meta
New York, New York, United States
Summary
🤩
Rockstar🎓
Top SchoolSebastian Ament is a Staff Research Scientist at Meta with six years of experience specializing in machine learning efficiency, AutoML, and Bayesian optimization, where he leads AutoML for Large Models to unlock significant compute and OPEX savings across recommendation and foundational models. He pairs foundational research—first‑author publications at NeurIPS and ICML—with production impact, translating novel surrogate modeling and optimization techniques into large‑scale training efficiency wins. A core contributor to BoTorch, he improved numerical stability of acquisition functions (introducing LogExpectedImprovement and related methods), strengthening a widely used open‑source Bayesian optimization stack. Based in New York, he also mentors and onboards researchers, drives cross‑functional roadmaps, and has applied ML to practical problems from AR/VR hardware simulation to low‑carbon concrete formulations for data centers.
6 years of coding experience
10 years of employment as a software developer
Bachelor of Arts (B.A.), Mathematics and Computer Science, Bachelor of Arts (B.A.), Mathematics and Computer Science at New York UniversityDoctor of Philosophy (PhD), Computer Science, Doctor of Philosophy (PhD), Computer Science at Cornell University
German, English
Github Skills (10)
pytorch10
machine-learning10
bayesian10
numerical-methods10
optimisation10
optimization10
python9
computer-engineering9
gpytorch7
unit-testing7
Programming languages (4)
Github contributions (5)
meta-pytorch/botorch
Aug 2022 - Dec 2022
Bayesian optimization in PyTorch
Role in this project:
ML Engineer / Software Engineer (focus on numerical methods and optimization)
ML Engineer / Software Engineer (focus on numerical methods and optimization)Contributions:29 reviews, 22 commits, 84 PRs in 4 months
Contributions summary:Sebastian made significant contributions to the `botorch` repository, focusing on improving numerical stability and expanding the functionality of Bayesian optimization tools. They introduced `LogExpectedImprovement` and `LogNoisyExpectedImprovement`, addressing numerical issues in the original EI implementations and enhancing gradient-based optimization. Additionally, they implemented `LogProbabilityOfImprovement`, ensuring improved acquisition function stability. Their work included refactoring and adapting existing code for deprecated functions as well as refactoring code to enhance the code's robustness and numerical stability.
bayesianpytorch
facebookresearch/SustainableConcrete
Oct 2023 - Jun 2026
Repository to track versions of concrete strength data, models, and active learning proposals.
Contributions:3 PRs, 69 pushes, 16 branches in 2 years 8 months
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.