Fitsum Reda
Principal Scientist at NVIDIA
California, United States
Summary
🤩
Rockstar🎓
Top SchoolFitsum Reda is a Principal Scientist at NVIDIA with nine years of experience advancing deep generative models, large language models, and vision-language systems from research to production. He holds a PhD in Electrical Engineering from Vanderbilt and an MS in Computer Vision and Robotics from Heriot-Watt, and has held research roles at Google and Siemens. His work spans LLMs, VLMs, vision tokenizers, and large video generation, and he has contributed practical improvements to high-profile open-source projects like FlowNet2 and a Google Research frame-interpolation implementation. Known for bridging research rigor with engineering pragmatism, he often tackles dataset pipelines, high-resolution image handling, and performance/compatibility fixes that make cutting-edge models more usable in practice. Based in California, he combines deep academic training with hands-on contributions that accelerate real-world generative AI workloads.
9 years of coding experience
5 years of employment as a software developer
Heriot-Watt University Edinburgh CampusDoctor of Philosophy (Ph.D.) Electrical Engineering, Doctor of Philosophy (Ph.D.) Electrical Engineering at Vanderbilt University
Github Skills (30)
pytorch10
python10
machine-learning10
deep-learning10
tensorflow10
optical-flow10
computer-vision10
image-processing9
mask-rcnn9
apache-beam9
faster-rcnn9
cuda8
dataprep8
data-loading8
data-preprocessing8
Programming languages (5)
Github contributions (5)
google-research/frame-interpolation
Feb 2022 - Jan 2023
FILM: Frame Interpolation for Large Motion, In ECCV 2022.
Role in this project:
ML Engineer
ML EngineerContributions:2 reviews, 48 commits, 7 PRs in 11 months
Contributions summary:Fitsum primarily contributed to the project by modifying code related to frame interpolation. They added support for various image formats, addressed padding issues in the interpolation process, and introduced features for high-resolution image handling. Furthermore, the user made improvements to the datasets pipeline, which involves the creation of TFRecord files, and improved the usability by adding a progress bar.
frameeccvoptical-flowfilmcomputer-vision
NVIDIA/flownet2-pytorch
Dec 2017 - Apr 2020
Pytorch implementation of FlowNet 2.0: Evolution of Optical Flow Estimation with Deep Networks
Role in this project:ML Engineer
ML EngineerContributions:65 commits, 14 PRs, 68 pushes in 2 years 4 months
Contributions summary:Fitsum made several updates to the `nvidia/flownet2-pytorch` repository, which implements optical flow estimation using deep networks. Their contributions include modifications to the datasets, losses, and main script, suggesting involvement in model training and evaluation. They addressed comments, updated configuration files, and adapted the code for PyTorch version compatibility and increased shared memory usage, demonstrating an understanding of the project's core functionality and operational considerations. They also implemented functionality for Resample2D.
pytorchdeep-networksevolutiondeep-learningoptical-flow
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.