Super stoked that UW SyFI (https://syfi.cs.washington.edu) members won a number of prizes at the MLSys’26 competition, NVIDIA Track. Hugre congrats to @KeisukeKamahori , @sudopowr , Yile Gu, Wei Shen, Steven Gao! Thanks to @nvidia , @modal , and the Flashinfer team for the support.

1st place in the GDN Track — Full-Agent Approach 2nd place in the GDN Track — Agent-Assisted Approach 3rd place in the DSA Track — Full-Agent Approach

SyFI Lab | SyFI Lab

Source: https://syfi.cs.washington.edu/ HomePeoplePublicationsTalksNewsBlogThe SyFI Lab at the University of Washington builds efficient and resilient infrastructure for the future of AI. As applications grow more complex, we bridge the gap between next-gen models and heterogeneous hardware through cross-stack innovation, delivering scalable, open-source systems validated by industrial partners.

Our research targets three key areas:

• Efficient AI: Optimizing algorithms and systems to maximize performance for training and inference.
• Flexible AI: Architecting systems that seamlessly adapt to diverse tasks, strategies, and model structures.
• Resilient AI: Ensuring AI system reliability at scale while leveraging AI to improve infrastructure robustness.

Blog Posts

Let AI Agents Write Your Serving Stack with VibeServe

May 12, 2026

We present VibeServe, a multi-agent system that synthesizes a complete LLM serving runtime end-to-end, specialized to a user-specified model, hardware, and workload.

SyFI in January 2026: A Big Month for Systems-Driven AI Research

January 31, 2026

January 2026 was a milestone month for the SyFI Lab, with six papers published across MLSys and ICLR—spanning inference, training, scheduling, retrieval, and model architecture.

Meet LLMc: Beating All Compression with LLMs

October 03, 2025

We present LLMc, an open-source tool to compress natural language using LLMs as the world’s most reference-packed dictionary.

Read more »

Talks

The Wrong Contract at Every Layer: Redesigning OS and Hardware for AI Datacenters

May 15, 2026Dimitrios Skarlatos — Carnegie Mellon University

Abstract

The AI datacenter stack is built on hardware-software contracts and abstractions that were never designed for the demands of the workloads they now serve. Memory systems strain under terabyte-scale capacity, heterogeneous AI accelerators have been forcefully pressed into deployment, and AI’s appetite for data increasingly collides with the privacy realities that infrastructure provides today. With datacenters projected to consume over 1,000 TWh annually, renegotiating the system and hardware stack is no longer optional. In this talk, I will share my research journey redesigning these contracts and where they lead. I will start with memory management, presenting Contiguitas and Learned Virtual Memory (LVM), a progression of OS and hardware redesigns that tackle virtual memory from fragmentation in production datacenters to near-ideal address translation. I will then discuss LithOS, the first operating system for efficient ML on GPUs, giving the OS the control over heterogeneous accelerators it has so far lacked. Finally, I will briefly describe how Cinnamon and Cerium bring these threads together to make encrypted AI practical at scale. I will conclude with open questions on how AI itself can help renegotiate the hardware-software contract, and where we go from here.

Speaker Bio

Dimitrios Skarlatos is an assistant professor in the Computer Science Department at Carnegie Mellon University. His research bridges computer architecture and operating systems with a focus on AI datacenter efficiency, privacy, and scalability. His work has been deployed in production datacenters and upstreamed into the Linux kernel. He has received the IEEE CS TCCA Young Computer Architect Award, the NSF CAREER Award, the Intel Rising Star Award, a Linux Foundation Faculty Award, an ISCA Best Paper Award, two ASPLOS Best Paper Awards, a CACM Research Highlight, four IEEE MICRO Top Picks, the joint ACM SIGARCH & IEEE CS TCCA Outstanding Dissertation Award, and over a dozen industry faculty awards from Meta, Intel, Amazon, Oracle, AMD, and VMware. His recent work led to the founding of LithosAI, a startup where he serves as the CEO.

Speaker Homepage »

Read more »