AI Systems
Building scalable, efficient and adaptable platforms for next-generation AI workloads, systems and hardware-aware optimization.
AI-Compiler
| Project Name: | UIUC PI: | IBM Tech Lead: |
| Minjia Zhang | Raghu Kiran Ganti, Mudhakar Srivatsa, Davis Wertheimer, Naigang Wang | |
| Maximizing the Efficiency of AI Chips with Compiler Techniques: From Tile-based Abstraction to Silicon | Jian Huang | Alaa Youssef, Chen Wang, Mudhakar Srivatsa, Raghu Kiran Ganti |
| Charith Mendis, Deming Chen, Sasa Misailovic | Viji Srinivasan, Prasanth Chatarasi, Mudhakar Srivatsa, Olivier Tardieu, Swagath Venkataramani |
Opti-Agents
| Project Name: | UIUC PI: | IBM PI: |
| Brighten Godfrey | Fabio Oliveira, Carlos Costa | |
| Klara Nahrstedt, Deming Chen | Hubertus Franke, Tamar Eilam, Asser Tantawi, Alaa Youssef, Eun K. Lee |
Distributed Inference
| Project Name: | UIUC PI: | IBM PI: |
| Network and Compute Co-design of a Distributed LLM Inference System Exploiting CXL and FPGA | Nam Sung Kim, Saksham Agarwal | Hubertus Franke, I-Hsin Chung |
| SLO-aware Serving Stack for Multi-Agent Inference | Fan Lai, Josep Torrellas | Hubertus Franke |
| Project: Enabling Elastic and SLO-Aware LLM Inference for Scalable and Cost-Effective Serving | Minjia Zhang, Tarek Abdelzaher | Fabio Oliveira, Carlos Costa |
Agent Systems
| Project Name: | UIUC PI: | IBM PI: |
| MASC: Building Reliability and Transaction Semantics for LLM Agents | Indranil Gupta | Rahul Krishna, Larissa Shwartz, Saurabh Jha, Chandra Narayanaswami |
| Distributed KV Cache Management and its Application to Multi-Agent Systems | Josep Torrellas, Tianyin Xu | Hubertus Franke, Fabio Oliveira |
| Agent KV: Co-Designing Agentic LLM Workloads and KV Cache Management for Efficient LLM Inference | Aishwarya Ganesan, Ramnatthan Alagappan | Fabio Oliveira, Hubertus Franke, Daniel Waddington |
AS2
| Project Name: | UIUC PI: | IBM PI: |
| Rapid Design Space Exploration, Synthesis, and Fabrication of SRAM-based In-memory Computing Architectures | Naresh Shanbhag | Gi-Joon Nam, David Kung |
| Toward Autonomous LLM-Driven SoC Generation with Verifiable Correctness, Coverage-Guided Testing, and PPA-Aware Optimization | Deming Chen | Gi-Joon Nam, David Kung |