Steven G

New multimodal model architectures shouldn't require new serving systems. Introducing our work, M* (M-Star): a universal serving system for multimodal models that separates what a model computes - a dataflow graph - from how it runs: placement, scheduling, batching, and transport. Joint work across @uwcse, @StanfordAILab, and @CMU_ECE with Atindra Jha, Naomi Sagan, Irmak Sivgin, Rohan Sanda, @ste_veng, Mark Horowitz, @LukeZettlemoyer, Olivia Hsu, @jure, @bariskasikci, and @thepadawang.

Modern multimodal models aren't a single decode loop anymore; they're composite. M* is one runtime that serves them all, and it matches or beats every specialized system: up to 2.7× on omni TTS, 12.5× on world-model rollouts. Learn more here: ai.stanford.edu/blog/mstar/

New distributed training strategies should not require new distributed runtimes. Introducing Piper: a programmable PyTorch training system for deploying complex training strategies by separating model placement and GPU scheduling from model code. 📄 arxiv.org/abs/2606.11169