​Join us for an evening devoted to efficient AI computing @ NeurIPS 2025, with a focus on making large language models more compute- and energy-efficient across the entire lifecycle: pre-training, post-training, reinforcement learning, inference, and systems design.

​The event will bring together faculty, researchers, and industry practitioners for a series of short technical talks and discussions on:

• ​Efficient pre-training – data, objectives, and architectures that improve sample, compute, and memory efficiency.

• ​Efficient post-training – instruction tuning, preference optimization, and distillation under strict efficiency constraints.

• ​RL for efficiency and control – reinforcement learning for policy optimization, routing, and adaptive resource-aware behavior.

• ​Inference-time optimization – sparsity, quantization, caching, and decoding strategies for low-latency, cost-effective serving.

• ​System and hardware efficiency – large-scale training/serving infrastructure and algorithm–systems–hardware co-design for end-to-end efficiency.

​We will discuss how efficiency, scalability, and co-design jointly shape the next generation of LLM training and deployment in both research and production environments. Following the talks, participants are invited to continue the conversation over a casual dinner, with space for deeper technical exchange, networking, and exploring future directions in efficient AI at scale.

​✨ Special Guests — Invited Speakers

​🎓 Jason Cong — Distinguished Professor @ UCLA and Member of the US National Academy of Engineering, pioneering researcher in electronic design automation, FPGA design, and domain-specific/customizable computing.

​🎓 Song Han — Associate Professor @ MIT Leading expert in efficient AI computing and hardware-aware machine learning.

​🎓 Zhijian Liu — Assistant Professor @ UCSD Researcher focused on efficient deep learning systems and algorithms.

​🎓 Tianlong Chen — Assistant Professor @ UNC Chapel Hill Specialist in optimization, sparse modeling, and efficient learning.

​🎓 Banghua Zhu — Assistant Professor @ UW; Principal Research Scientist @ NVIDIA; Prior co-founder at Nexusflow AI; working on RL

​🎓 Fangyu Liu — Staff Research Scientist @ Google DeepMind working on Gemini pretraining

​🎓 Yuchen Zhuang — Research Scientist @ Google DeepMind focusing on RL and post-training

​🎓 Ryan Hanrui Wang — CEO @ Eigen AI Bridging the gap between academic research and industrial AI infrastructure.

​…More invited speakers to be announced 🔜

​Your Host: Eigen AI

​Eigen AI is a full-stack AI infrastructure platform built to make modern AI systems faster, lighter, and truly scalable.

​From data generation → training → compression → acceleration → deployment, Eigen AI enables teams to ship high-performance systems at a fraction of the cost, latency, and complexity.

​We believe the next leap in AI will be defined by efficiency at scale — and we’re building the infrastructure that makes it possible.