Amir Bar
Research Scientist at Meta
New York, New York, United States
Summary
🤩
Rockstar🎓
Top SchoolAmir Bar is a research scientist at Meta with 11 years of experience building and deploying machine learning systems, particularly in computer vision and medical imaging. He progressed from research intern to postdoc and now staff researcher, combining academic rigor from a PhD track with hands-on engineering across startups and large tech. At Zebra Medical Vision he led AI research and tech teams, delivering clinical-grade models, and his open-source contributions include practical enhancements to a popular Mask R-CNN TensorFlow implementation for better visualization and numerical stability. Based in New York, he focuses on interpretability and robust training pipelines and is explicitly not seeking quant/finance roles. Colleagues know him for bridging research and production, improving debuggability in complex vision models while keeping model behavior grounded in real-world preprocessing changes.
11 years of coding experience
6 years of employment as a software developer
Doctor of Philosophy - PhD Computer Science, Doctor of Philosophy - PhD Computer Science at University of California, BerkeleyDoctor of Philosophy - PhD Computer Science, Doctor of Philosophy - PhD Computer Science at Tel Aviv University
Hebrew, English
Stackoverflow
839reputation
43kreached
24answers
2questions
tensorflow
top-5%
Github Skills (16)
computer-vision10
machine-learning10
tensorflow10
python10
segmentation9
image-segmentation9
data-visualisation9
data-visualization9
data-visualizations9
lua6
keras6
http6
neural-network6
convolution6
deconvolution6
Programming languages (3)
Github contributions (5)
CharlesShang/FastMaskRCNN
Apr 2017 - Jul 2017
Mask RCNN in TensorFlow
Role in this project:
ML Engineer
ML EngineerContributions:12 commits, 13 PRs, 4 pushes in 2 months
Contributions summary:Amir focused on enhancing the model's capabilities by adding input visualization to the training pipeline, including ground truth bounding boxes and masks. They added summaries for predictions like RPN bounding boxes and final predicted bounding boxes and masks to improve the model's interpretability and debugging capabilities. The user also corrected parameters passed to the network to reflect changes made during preprocessing steps. They also made adjustments to calculations within the bounding box transformation to improve numerical stability.
maskrcnntensorflowmask-rcnn
amirbar/DETReg
Jun 2021 - Apr 2022
Official implementation of the CVPR 2022 paper "DETReg: Unsupervised Pretraining with Region Priors for Object Detection".
Contributions:1 release, 2 reviews, 44 commits in 10 months
pytorchunsupervised-learningdeep-learningunsupervisedobject-detection
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.