Employment
Employment history is unavailable.
Education
Education history is unavailable.
Projects & Funding
Projects & funding information is unavailable.
Publications (16)
- BigMac: Breaking the Pareto Frontier of Compute and Memory in Multimodal LLM Training Save
- ExpertPlex: A High-Goodput Disaggregated Serving System for MoE LLMs with Adaptive Persistent Kernels Save
- Heddle: A Distributed Orchestration System for Agentic RL Rollout Save
- PROBE: Co-Balancing Computation and Communication in MoE Inference via Real-Time Predictive Prefetching Save
- ReLibra: Routing-Replay-Guided Load Balancing for MoE Training in Reinforcement Learning Save
- ScalingCache: Extreme Acceleration of DiTs through Difference Scaling and Dynamic Interval Caching Save
- UltraEP: Unleash MoE Training and Inference on Rack-Scale Nodes with Near-Optimal Load Balancing Save
- xHC: Expanded Hyper-Connections Save
- SlimPipe: Memory-Thrifty and Efficient Pipeline Parallelism for Long-Context LLM Training Save
- Kling-omni technical report Save