Sam Pepose
Founder at Antron
San Francisco, California, United States
Summary
🤩
Rockstar🎓
Top SchoolSam Pepose is a founder and machine learning engineer with 13 years of experience building research-driven computer vision and generative modeling systems from research to product. Formerly an Engineering Manager and Applied Research Scientist at Facebook, he led and supported 0-to-1 projects in NeRFs, novel view synthesis, large-scale unsupervised learning, and photorealistic avatar generation. He contributes to high-profile open-source projects such as Relay and Detectron2, where he improved mutation handling, deferred fragments, and robustness in object detection code paths—work that signals an attention to both algorithmic detail and production reliability. Based in San Francisco and grounded in CS and applied math from Vanderbilt (with further CS study at Stanford), he now builds the data engine for physical intelligence at Antron, blending deep research experience with practical system design. An understated strength is his track record of refining low-level model and data plumbing—the kind of work that makes advanced ML systems reliably deployable.
13 years of coding experience
9 years of employment as a software developer
Whitfield SchoolBachelor of Science (B.S.), Computer Science, Bachelor of Science (B.S.), Computer Science at Vanderbilt UniversityComputer Science, Computer Science at Stanford University
Spanish, English
Stackoverflow
1,821reputation
295kreached
52answers
17questions
Github Skills (29)
pytorch10
javascript10
python10
jes10
machine-learning10
wpgraphql10
relayer10
image-segmentation10
segmentation10
jestjs10
wordpress-graphql10
jtest10
graphql10
object-detection10
computer-vision10
Programming languages (10)
Github contributions (5)
facebookresearch/detectron2
Oct 2019 - Feb 2020
Detectron2 is a platform for object detection, segmentation and other visual recognition tasks.
Role in this project:
ML Engineer
ML EngineerContributions:7 commits, 1 comment in 3 months
Contributions summary:Sam primarily contributed to the `detectron2` repository by addressing issues related to object detection and image segmentation. Their work included fixing shape inference for the `ConvTranspose2d` layer, enhancing file I/O buffering, and ensuring the consistent device usage after box mode conversions, which improved the robustness of the library. Furthermore, the user refactored code to enhance its usability and stability. The user also added type hints and made box mode JSON serializable, improving the project's maintainability.
detectron2deep-learningface-recognitionrecognitionobject-detection
facebook/relay
May 2016 - Jul 2016
Relay is a JavaScript framework for building data-driven React applications.
Role in this project:Full-stack Developer
Full-stack DeveloperContributions:9 commits, 6 comments, 2 issues in 2 months
Contributions summary:Sam made significant contributions to the Relay JavaScript framework, focusing on mutation handling and data fetching logic. They implemented features such as range delete mutations, which involved modifying record stores and update payloads. The user also worked on deferred fragments, ensuring correct storage and resolution. Additionally, they contributed to the test suite, demonstrating a focus on code quality and validation.
data-drivenreactrelayreact-applicationsjavascript-framework
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.