Shaoting Feng

DroidSpeak turbocharges multi-LLM pipelines by sharing KV caches across fine-tuned models, slashing latency by up to 2.6× and boosting throughput by 3× with negligible accuracy loss.

RAGServe turbocharges RAG by pruning and adapting per-query configurations on the fly, slashing latency by up to 2.5× without sacrificing quality.

Scalable, bandwidth-preserving algorithms for AFDX virtual links that optimise message allocations.

GIPUT models objects with realistic shapes, enabling UAVs to learn optimal trajectories while considering photo coverage, energy consumption, and bandwidth utilization. It achieves twice the efficiency of state-of-the-art algorithms.

The first open-source Knowledge Delivery Network (KDN) that accelerates LLM applications up to 8x faster, at 8x lower cost.

Scale from single vLLM instance to distributed vLLM deployment without changing any application code.

Systematic and comprehensive benchmarks for LLM systems.

Developed and implemented a robust MAC layer for the LoRa communication protocol, optimizing for long-range, low-power, and anti-interference performance. Designed a complete software stack integrating MCU control and advanced features like timeout retransmission and duty cycle sleep management for a software-defined radio platform.