Yi (Jerry) Li
I am a Machine Learning Researcher at Qualcomm. I work on efficient and reliable multimodal AI, spanning video understanding, vision-language models, robotics, and physical AI.
I received my PhD in Electrical and Computer Engineering from UC San Diego in 2024, advised by Nuno Vasconcelos, and my BEng in Electronic Engineering from CUHK in 2017. I previously interned at the MIT-IBM Watson AI Lab and Intel AI Lab.
Outside research, I enjoy music, play the piano, and am learning the flute. My current favorites are Beethoven’s String Quartet, Op. 132 and Sibelius’s Symphony No. 6.
Research interests
Multimodal & Agentic AI
- Vision-language models
- Video understanding
- Multimodal agents
Efficiency & Reliability
- Efficient inference
- Model reliability
- Privacy & fairness
Robotics & Physical AI
- Spatial reasoning
- Memory
- Embodied intelligence
Publications
Latent Bridge: Feature Delta Prediction for Efficient Dual-System Vision-Language-Action Model Inference
TL;DR
Predicting changes in visual-language features to reduce the cost of VLA inference.
No publications match your search. Try another title, author, or topic.
* Equal contribution
News
CoMET accepted to the NeurIPS Evaluations & Datasets track.
ForeSea accepted to ECCV.
Memory-augmented selective prediction accepted to ICLR.
EgoPrivacy accepted to ICML.
Defended my PhD thesis, Holistic Bias Mitigation in Computer Vision and Beyond. Watch the defense
Earlier milestones
Attended the CVPR Doctoral Consortium, mentored by Kate Saenko.
Named a CVPR Outstanding Reviewer.
SViTT accepted to CVPR 2023.