Yan Kang

Senior Research Engineer

Seattle

Yan is a Senior Research Engineer in the Research Engineering and Design Lab at Adobe Research. He spearheads efforts to streamline computational costs across Adobe’s models at various scales, by leveraging sparse MoE, Neural Architecture Search, pruning, distillation, block caching, approximated Attention, Attention KV compression, post-training inference optimizatoin, etc. To push efficiency to its limits, Yan also works on optimizing the heavy cloud-serving models for on-device deployment, enabling high-performance solutions on edge devices. He is a technical lead and works with a group of research engineers and machine learning engineers to conduct model efficiency efforts.

Contact: yankang [at] adobe [dot] com

News

  • Nov. 2024. I’m hiring multiple interns for different topics related to GenAI efficiency. 
  • Sep. 2024. Won Adobe Tech Excellence Award, among few recipients worldwide.

Tech Transfers

  • Firefly 1.0
  • Firefly 2.0
  • GenerativeFill 1.0
  • GenerativeFill 3.0
  • MetaCAF (Content Aware Fill)
  • SuperCAF (Content Aware Fill)
  • Wire segmentation
  • Panoptic segmentation
  • Select person
  • On-device image inpainting
  • On-device image super-resolution
  • On-device video super-resolution
  • In-browser panoptic segmentation

Previous graduated interns

Publications

Mixture of Efficient Diffusion Experts Through Automatic Interval and Sub-Network Selection

Ganjdanesh, A., Kang, Y., Liu, Y., Zhang, R., Lin, Z., Huang, H. (Oct. 1, 2024)

European Conference on Computer Vision (ECCV)

SNED: Superposition Network Architecture Search for Efficient Video Diffusion Model

Li, Z., Kang, Y., Liu, Y., Liu, D., Hinz, T., Liu, F., Wang, Y. (Jun. 19, 2024)

CVPR 2024

AT-EDM: Attention-Driven Training-Free Efficiency Enhancement of Diffusion Models

Wang, H., Liu, D., Kang, Y., Li, Y., Lin, Z., Jha, N., Liu, Y. (Jun. 17, 2024)

CVPR 2024