Shivam Shrirao
Research Scientist at Bria AI
India
Summary
🤩
Rockstar🎓
Top SchoolShivam Shrirao is a research scientist with 7 years of hands-on experience building and scaling state-of-the-art deep learning systems, particularly in diffusion-based image generation, computer vision, and large-scale training pipelines. He has led AI teams and shipped production-ready solutions—from photorealistic virtual try-on and NeRF-based view synthesis to high-throughput distributed training for 24B-parameter diffusion models. Shivam pairs strong applied ML skills with deep systems expertise (Linux, networking, binary exploitation, and reverse engineering), enabling memory- and latency-optimized workflows such as fast attention and TensorRT-accelerated inference. A contributor to the Hugging Face Diffusers ecosystem, he implemented fast attention and xformers integration to speed up and reduce memory use in generative models. He’s known for squeezing large models into practical production constraints (e.g., cutting stable-diffusion fine-tuning VRAM needs dramatically) while exploring security and low-level performance tradeoffs. Based in India, he blends research curiosity with product-focused engineering to deliver creative, scalable AI features.
7 years of coding experience
3 years of employment as a software developer
B. Tech Computer Science, B. Tech Computer Science at MIT World Peace University
Github Skills (6)
attention-mechanism10
diffusion-models10
computer-vision10
pytorch10
machine-learning10
python10
Programming languages (5)
Github contributions (5)
ShivamShrirao/diffusers
Sep 2022 - Jan 2023
🤗 Diffusers: State-of-the-art diffusion models for image and audio generation in PyTorch
Role in this project:
ML Engineer
ML EngineerContributions:1 review, 127 commits, 20 PRs in 3 months
Contributions summary:Shivam primarily focused on improving the performance and functionality of diffusion models within the Diffusers library. The user added fast attention mechanisms to the attention layers, resulting in significant speedups and reduced memory requirements. They also worked on the integration of xformers for memory-efficient attention and the removal of NSFW filters. Furthermore, they worked on training scripts related to the dreambooth method.
pytorchartdeep-learningstate-of-the-artaudio
ShivamShrirao/ntt_pollution_hybrid_lstm_dnn
Mar 2019 - Sep 2020
Contributions:16 commits, 3 pushes in 1 year 6 months
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.