Prathik Rao
Software Engineer at Microsoft
Menlo Park, California, United States
Summary
🤩
Rockstar🎓
Top SchoolPrathik Rao is a software engineer with eight years of experience specializing in accelerating transformer-based vision, diffusion, and language models, currently contributing to ONNX Runtime Training at Microsoft from Menlo Park. He has practical experience productionizing ML workflows—integrating ONNX Runtime into Hugging Face's widely used diffusers library and adding training-grade features to ONNX Runtime such as gradients, model weight loading, and BFloat16 support. Earlier work includes scaling Azure Resource Manager to handle billions of daily requests and building model-serving SDKs and tooling for computer vision startups and research labs. Trained in computer engineering and computer science at USC with a graduate certificate in AI from Stanford, he combines systems-level engineering with hands-on ML model optimization. A less obvious strength is his track record of moving research and prototypes into robust, debuggable pipelines—from teaching and academic research to shipping production training performance improvements.
8 years of coding experience
3 years of employment as a software developer
Bachelor of Science - BS, Computer Engineering and Computer Science, Bachelor of Science - BS, Computer Engineering and Computer Science at University of Southern CaliforniaGraduate Certificate, Artificial Intelligence, Graduate Certificate, Artificial Intelligence at Stanford University
Github Skills (19)
pytorch10
python10
diffusion-models10
machine-learning10
onnx10
hardware-acceleration10
diffusers10
diffusion-probabilistic-models10
ml10
mle10
diffusion-probabilistic10
deep-learning10
image-generation10
onnxruntime10
model-optimization10
Programming languages (3)
Github contributions (5)
microsoft/onnxruntime
Sep 2022 - Oct 2022
ONNX Runtime: cross-platform, high performance ML inferencing and training accelerator
Role in this project:
ML Engineer
ML EngineerContributions:77 reviews, 3 commits, 132 PRs in 22 days
Contributions summary:Prathik's commits primarily focus on enhancing the ONNX Runtime, an ML inferencing and training accelerator. They implemented gradients for mathematical functions like sin and cos within the training module, improving performance. Furthermore, they added features to load pre-trained model weights, supporting the use of models like Whisper and T5. Additionally, the user worked on caching exported ONNX models in ORTModule, and added support for BFloat16 in various operators, indicating a focus on optimizing performance and expanding compatibility for training.
runtimetrainingtensorflowai-frameworkaccelerator
huggingface/diffusers
Nov 2022 - Dec 2022
🤗 Diffusers: State-of-the-art diffusion models for image, video, and audio generation in PyTorch and FLAX.
Role in this project:ML Engineer
ML EngineerContributions:8 reviews, 5 commits, 23 PRs in 1 month
Contributions summary:Prathik contributed to the integration of ONNX Runtime (ORT) for model optimization within the `diffusers` repository, focusing on image generation tasks. They worked on adapting training scripts to leverage ORT, specifically in the context of unconditional image generation. Their commits include code modifications to incorporate ORT, along with bug fixes and updates to the training pipeline.
pytorchartdeep-learningimage2imagestate-of-the-art
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.