Rex Li
contact: silicontalks@gmail.com
Education Background
- PhD in Electrical Engineering, UCLA (2016)
Industry Experience
- Research Scientist, Novumind (2017 - 2019)
- Microarchitecture for AI accelerator
- Hardware-aware model optimization
- AI Research Scientist Manager, Facebook (2019 - Present)
- Hardware-model co-design and optimization
- Multimodal models for on-device Contextual AI
- Lead effort on-device pillar of PyTorchVideo, enabled first on-phone realtime video understanding model in community
- Generative AI for emerging user experiences
Selected Talks, Publications and Patents
- Gong, Mohan, Dhingra, Bazin, Li, et al., MMG-Ego4D: Multi-Modal Generalization in Egocentric Action Recognition, to appear in CVPR 2023
- PyTorch Video Accelerator, ICCV 2021 Efficient Video Understanding Tutorial
- Fan, Murrel, Wang, Alwala, Li, Li et al., PyTorchVideo: A Deep Learning Library for Video Understanding, ACM Conference on Multimedia, 2021
- Li, Li et al., Improving Efficiency in Neural Network Accelerator using Operands Hamming Distance Optimization, Neurips 2019 EMC2 Workshop
- Du, Du, Li et al., A Reconfigurable Streaming Deep Convolutional Neural Network Accelerator for Internet of Things, IEEE Transactions on Circuits and Systems I: Regular Papers, 2018. (Winner of IEEE Darlington Best Paper Award 2021)
- Convolution operation device and convolution operation method, US20180137414A1
- Convolution operation device and method, US10169295B2
- Method and system for elastic precision enhancement using dynamic shifting in neural networks, US20200050429A1
- Buffer device and convolution operation device and method, US10162799B2
- Multi-layer neural network, US10552732B2
- Neural network based on total hamming distance, US11507814B1
- Dynamically Selecting Neural Networks for Detecting Predetermined Features, US20220101098A1
- Camera control using system sensor data, US20230012426A1