Kevin Li
Foundation Model Research at Adobe
- Role
- Applied Research Scientist at Adobe
- Location
- Berkeley, CA, US
- LinkedIn followers
- 500 followers
About Kevin Li
Building multimodal foundation model at Adobe Firefly. https://thekevinli.github.io/
Experience
Applied Research Scientist
Aug 2023 — Present · CA, US
Co-lead of multimodal pretraining for Firefly Image 5 Preview, demoed at Adobe MAX 2025.Image 5 offers 4MP text-to-image, instruction editing, and layered generation/editing. I led ablation studies on the architecture and data combinations, and drove decisions on the final production editing model\'s training recipe, scaling to distributed training.Recently released tech report - UniFusion, the first architecture to use only VLM as unified multimodal encoder without auxiliary signals from VAE to do image generation & editing with competitive quality, beating Flux.1 [dev] and Bagel, respectively.Other products that I have led or been one of the core contributors- Firefly Image 4 model pre-training architecture design and post-training (aesthetic finetuning)- Firefly Image 3 model personalization- Firefly Image 2 style transfer
Education
University of California, Berkeley
Bachelor of Science - BS, Electrical Engineering and Computer Science
2019 — 2023
Mingdao High School
High School, Mathematics Gifted Program
2013 — 2019
Skills
- Web Development
- Negotiation
- 研究
- 社交媒体
- Software Engineers
- Bioinformatics
- 推销
- 领导力
Find verified contacts for anyone on LinkedIn
Unifers gives sales teams verified emails and direct dials, enriched profiles, and outreach that lands in the inbox.
Free plan included · No credit card required
This profile is compiled from publicly available professional sources. Unifers is not affiliated with or endorsed by LinkedIn. Request removal of this profile.