Yi Wang
Video Team Leader at Intel Corporation
Shanghai, China
Summary
🤩
Rockstar🎓
Top SchoolYi Wang is a video technology leader with 14 years of experience designing and shipping video codecs, multimedia frameworks, and real-time conferencing systems across Intel, Cisco, and Marvell. He combines deep research background in video compression (PhD-level training) with hands-on embedded and Android integration work—having implemented H.264/AVC codecs, OpenMax/GStreamer components, and platform-specific DRM and hardware codec solutions. At Intel he leads video teams building IP- and platform-level solutions, and his prior work includes integrating codecs into WebRTC and MCU architectures for telepresence. In recent years he has also contributed to high-profile open-source ML infrastructure (Hugging Face repos) improving distributed training, optimization workflows and CPU/Intel XPU performance, reflecting a strong cross-domain systems focus. Known for architecting media software and APIs end-to-end, he bridges low-level C/C++/assembly optimization with Python-based ML and backend tooling to accelerate real-time media and inference pipelines.
14 years of coding experience
8 years of employment as a software developer
Doctor, Electronic Engineering and Information Science, Doctor, Electronic Engineering and Information Science at University of Science and Technology of China
Github Skills (37)
transformers10
pytorch10
distributed-training10
acc10
intel10
python10
acceleration10
machine-learning10
accelerator10
inference10
reinforcement-learning10
ml10
mle10
transformer-models10
huggingface10
Programming languages (5)
Github contributions (5)
huggingface/text-generation-inference
Sep 2023 - Mar 2025
Large Language Model Text Generation Inference
Role in this project:
Back-end Developer & ML Engineer
Back-end Developer & ML EngineerContributions:25 reviews, 64 PRs, 189 comments in 1 year 6 months
Contributions summary:Yi primarily contributed to the integration of Intel XPU support within the text generation inference framework. Their work involved modifying existing code to enable bfloat16 support for CPU, and adding functionality for Intel XPU, as well as CPU support. They also addressed issues related to model loading and potential crashes, suggesting a focus on performance optimization and platform compatibility within the context of large language models. Furthermore, they integrated prefill chunking, prefix caching and Flash Decoding kernels to accelerate inference.
nlppytorchlanguage-modelbloombert
huggingface/optimum-intel
Aug 2022 - Jan 2023
🤗 Optimum Intel: Accelerate inference with Intel optimization tools
Role in this project:ML Engineer
ML EngineerContributions:17 reviews, 8 commits, 32 PRs in 5 months
Contributions summary:Yi primarily contributed to the optimization and improvement of the Intel-specific components within the `optimum-intel` repository. Their work focused on addressing issues related to loss calculation and import errors within the neural compressor, specifically when the loss weights were set to zero. They also implemented and integrated updates from the transformers library regarding download and caching mechanisms, and aligned the caching and downloading of configurations. Additionally, they addressed performance considerations by disabling the amp backend for certain quantized models and adding a speed metric to the trainer.
pytorchtransformersdeep-learningpruningoptimization
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.