Our research on Edge Intelligence focuses on bringing AI capabilities closer to end users and devices. We investigate efficient DNN and LLM inference at the network edge, federated learning for privacy-preserving distributed training, and cloud-edge-end collaborative computing architectures. Our goal is to enable intelligent, low-latency services while optimizing resource utilization across heterogeneous edge networks.

Key topics include:

• Mobile edge computing and caching
• Federated learning and continual learning
• Speculative decoding for LLM inference
• Green edge intelligence and energy-efficient AI
• Stochastic network optimization