Tyler Griggs

Tyler Griggs

I am a third year PhD student in Computer Science in the UC Berkeley Sky Computing Lab advised by Ion Stoica and Matei Zaharia. Previously, I worked in Network Infrastructure at Google Cloud. Before that, I graduated from Harvard with a BA in Computer Science advised by James Mickens.

My research interests are in designing and building flexible and efficient systems for model training and inference, and leveraging these systems to develop models useful in real-world tasks. My current research focuses on systems for post-training, especially reinforcement learning. Along with several wonderful collaborators, I am a co-lead of the NovaSky team. Our work has been featured in The New York Times, The Wall Street Journal, and The Information.

Current Projects

SkyRL
SkyRL: A Modular Full-stack RL Library for LLMs GitHub stars
SkyRL is a modular, performant RL framework built for multi-turn agentic training, using a simple and flexible design that allows both high-performance execution and adaptation to a wide variety of training scenarios.
Train your agent now with SkyRL, and don't hesitate to reach out!
[GitHub] [Blog]
SkyThought
SkyThought: A repository to reproduce NovaSky's research GitHub stars
The SkyThought repository holds all code needed to reproduce all prior work of the NovaSky team, such as efficient reasoning, reinforcement learning, test-time scaling, and more. See all our work on our website.
[GitHub]

Prior Projects

Reasoning models can be effective without thinking
Reasoning models can be effective without thinking
Wenjie Ma, Jingxuan He, Charlie Snell, Tyler Griggs, Sewon Min, Matei Zaharia
arXiv preprint [Paper]
Structure not Content
LLMs Can Easily Learn to Reason from Demonstrations: Structure, not content, is what matters!
Dacheng Li*, Shiyi Cao*, Tyler Griggs*, Shu Liu*, Xiangxi Mo, Eric Tang, Sumanth Hegde, Kourosh Hakhamaneshi, Shishir G Patil, Matei Zaharia, Joseph E Gonzalez, Ion Stoica
arXiv preprint [Paper]
MoE-Lightning
MoE-Lightning: High-Throughput MoE Inference on Memory-constrained GPUs
Shiyi Cao, Shu Liu, Tyler Griggs, Peter Schafhalter, Xiaoxuan Liu, Ying Sheng, Joseph E Gonzalez, Matei Zaharia, Ion Stoica
ASPLOS 2025 [Paper]
SkyServe
SkyServe: Serving AI Models across Regions and Clouds with Spot Instances
Ziming Mao, Tian Xia, Zhanghao Wu, Wei-Lin Chiang, Tyler Griggs, Romil Bhardwaj, Zongheng Yang, Scott Shenker, Ion Stoica
EuroSys 2025 [Paper] [Code]
Mélange
Mélange: Cost Efficient Large Language Model Serving by Exploiting GPU Heterogeneity
Tyler Griggs, Xiaoxuan Liu, Jiaxiang Yu, Doyoung Kim, Wei-Lin Chiang, Alvin Cheung, Ion Stoica
arXiv preprint [Paper] [Code]