Catalyst: Catalyst Research Summit 2026

CMU Catalyst Research Summit

March 23, 2026 · Rm 6115, Gates and Hillman Centers, Carnegie Mellon University, 4902 Forbes Ave, Pittsburgh

About

The Catalyst Lab is delighted to invite you to the CMU Catalyst Research Summit, which will take place on March 23 on the CMU campus in Pittsburgh. At the summit, the lab will showcase its ongoing research on agentic systems, ML compilers, and multimodal AI systems. The program will bring together Catalyst faculty, students, and collaborators working on core systems and infrastructure challenges in modern AI. It will feature research talks highlighting frontier work from Catalyst, along with panel discussions and poster sessions designed to foster in-depth interaction with faculty and students. The Catalyst Lab will also share the roadmap for several of its ongoing open-source projects.

Schedule

Time	Event
8:30 AM – 9:30 AM	Breakfast
9:30 AM – 11:00 AM	Catalyst Research Keynote
11:00 AM – 11:30 AM	Coffee Break
11:30 AM – 12:30 PM	Catalyst Project Presentations (15 min per talk) Streaming Tensor Programs (Olivia Hsu) FlashRT: Realtime and Interactive Multimodal Generation (Krish Agarwal) Core Attention Disaggregation in LM Training (Yonghao Zhuang) FlexLLM: Token-Level Co-Serving of LLM Inference and Finetuning with SLO Guarantees (Gabriele Oliaro) TIRX and Event Tensor: Bringing ML Compiler for Frontier Kernel Programming (Bohan Hou, Hongyi Jin)
12:30 PM – 2:00 PM	Lunch Break
2:00 PM – 3:30 PM	Catalyst Project Presentations (15 min per talk) Bitsandbytes 2: Efficient Agent Training and Inference (Tim Dettmers) XGrammar-2: Efficient Dynamic Structured Generation Engine for Agentic LLMs (Yixin Dong) PithTrain: Efficient, Python-Native MoE Training in ~10K Lines of Code (Ruihang Lai, Hao Kang) Unlocking the Potential of Post-training and Model Architecture Codesign (Zhuoming Chen) Mirage Persistent Kernel: A Compiler and Runtime for Mega-Kernelizing Tensor Programs (Xinhao Cheng) Lithos: An Operating System for Efficient Machine Learning on GPUs (Eliot Solomon)
3:30 PM – 4:00 PM	Coffee Break
4:00 PM – 4:30 PM	Invited Lightning Talks (5 min per team)
4:30 PM – 5:30 PM	Panel Discussion: Co-Evolving Intelligence: MLSys × AI Agents in the Post-AGI Era (Moderator: Prof. Beidi Chen) Vinod Grover (NVIDIA) Fredrik Kjoelstad (Stanford) Gennady Pekhimenko (University of Toronto/NVIDIA) Yida Wang (Amazon) Minjia Zhang (UIUC)
5:30 PM – 8:00 PM	Poster Session with Reception — NSH Atrium

Parking

Gates Hillman Building address: 4902 Forbes Ave, Pittsburgh, PA 15213

For parking garage options please visit CMU Visitor Parking. This will have rates and locations of the parking garages.

East Campus Garage – 6 min walk to Gates Hillman Building. You will enter on the 5th floor. Elevators and stairs can be taken to the 6th floor.
Gates Garage – This garage is in the basement of the Gates Hillman Building. You will enter on the 1st–3rd floors. Elevators and stairs can be taken to the 6th floor.
Mehrabian Collaborative Innovation Center (CIC) Garage – Take the elevator to the Lobby of the Collaborative Innovation Center. Walk from CIC building to Gates Hillman Building. You will enter on the 3rd floor.
Dithridge Street Garage – 10 min walk to Gates Building.
Morewood Lot – 3 min walk to Gates Hillman Building. You will enter on the 4th floor. Walk to back elevators and take to 6th floor.

CMU campus parking map showing locations of East Campus Garage, Gates Garage, CIC Garage, and Morewood Lot relative to Gates Hillman Building