Sait Cakmak
Senior Research Scientist at Meta
New York, New York, United States
Summary
🤩
Rockstar🎓
Top SchoolSait Cakmak is a Senior Research Scientist at Meta’s Adaptive Experimentation group with eight years of experience applying Bayesian optimization and Gaussian process methods to real-world product problems. He earned a PhD in Operations Research from Georgia Tech, where his work focused on risk-averse optimization for black-box and simulation-based objectives. At Meta he helps maintain and improve prominent open-source tools like Ax and BoTorch, contributing backend refactors and performance improvements to widely used Bayesian optimization infrastructure. His open-source contributions to gpytorch and BoTorch include numerical robustness fixes, batch qMC sampling support, and optimizations that reduce computational cost in practice. Comfortable moving between research and production, he combines rigorous stochastic optimization theory with pragmatic software engineering to deliver scalable experimentation tooling. An interesting detail: he has a track record of improving core library internals (e.g., likelihood noise control and variational distributions) rather than just surface features, reflecting deep familiarity with probabilistic modeling codebases.
8 years of coding experience
6 years of employment as a software developer
Visiting Student, Economics, GPA: 4.00/4.00, Visiting Student, Economics, GPA: 4.00/4.00 at University of California, BerkeleyExchange Student, Economics, GPA: 3.93/4.00, Dean's Honor List, Exchange Student, Economics, GPA: 3.93/4.00, Dean's Honor List at University of California, IrvineBachelor of Science, Industrial Engineering; Economics, GPA: 3.91/4.00, Bachelor of Science, Industrial Engineering; Economics, GPA: 3.91/4.00 at Koç ÜniversitesiExchange Student, School of Business, Mikkeli, GPA: 4.40/5.00, Exchange Student, School of Business, Mikkeli, GPA: 4.40/5.00 at Aalto UniversityDoctor of Philosophy - PhD, Operations Research, Doctor of Philosophy - PhD, Operations Research at Georgia Institute of Technology
Turkish, İngilizce
Stackoverflow
1reputation
0reached
0answers
0questions
Github Skills (20)
pytorch10
python10
testing10
machine-learning10
multi-objective-optimization10
ml10
gaussian-processes10
mle10
bayesian10
optimisation10
gpu-acceleration10
gpytorch10
optimization10
model-driven9
model-optimization9
Programming languages (5)
Github contributions (5)
pytorch/botorch
Aug 2020 - Jan 2023
Bayesian optimization in PyTorch
Role in this project:
Back-end Developer / ML Engineer
Back-end Developer / ML EngineerContributions:17 releases, 253 reviews, 155 commits in 2 years 6 months
Contributions summary:Sait implemented support for batch qMC sampling, which is beneficial for optimizing batch GP models and acquisition functions. They implemented a method to evaluate qKG values of candidates through solving the inner optimization problem. The user also made modifications to existing functions, such as `_expand_bounds`, to work with greater-than-2-dimensional bounds to extend optimization capabilities. Further contributions include code adjustments, such as fixing a dtype bug and implementing checks for NaN values.
pytorchoptimizationmultiobjective-optimizationmachine-learningbayesian-optimization
facebook/Ax
Aug 2021 - Jan 2023
Adaptive Experimentation Platform
Role in this project:Back-end Developer
Back-end DeveloperContributions:1 release, 38 reviews, 73 commits in 1 year 5 months
Contributions summary:Sait's contributions primarily involved refactoring and improving the Ax experimentation platform. They removed deprecated code related to model transformations within the fully_bayesian module and fixed pyre errors and test failures in the botorch moo model. Further improvements were made to reduce the cost of initial condition generation in the fast_botorch_optimize function. Also, the user worked on supporting BoTorch input and outcome transforms in MBM.
experimentationadaptivesimulationadaptive-experimentation-platformexperimentation-platform
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.