U Lab @ UIUC
U Lab is an academic research lab led by Prof. Jiaxuan You at UIUC. Its mission is to accelerate knowledge discovery and sharing by architecting AI agents and supporting infrastructure that automate research and other knowledge processes while serving as trustworthy partners for people.
At a Glance
- Academic and industrial AI/ML researchers
- Developers building multi-LLM applications
- Teams optimizing response quality and inference cost
- Research-agent and scientific-computing developers
- +1 more
AI Tools by U Lab @ UIUC
(1)
LLMRouter
Open Source LLM Router Library
Discussions
No discussions yet
Be the first to start a discussion about U Lab @ UIUC
Latest News
LLMRouter paper submitted to arXiv, introducing unified infrastructure and the xRouteBench benchmark for LLM routing.
LLMRouter website published documentation describing the unified routing formulation, benchmark, deployment layer and 14.6% relative improvement over the strongest fixed-model baseline.
TinyScientist research-agent framework listed as an EMNLP 2025 publication.
PersonalizedRouter listed as a TMLR 2025 publication on personalized LLM routing via graph-based user-preference modeling.
Products & Services
An open-source foundation for selecting the right LLM for each request. It unifies data construction, router training, inference, evaluation and deployment, and includes more than 16 representative routers across single-turn, multi-turn and personalized routing.
A multi-scenario benchmark and routing-supervision pipeline spanning generic LLM tasks, memory-augmented reasoning, vision and video understanding, time-series and personalized routing; it records task quality and token-level inference cost under a shared protocol.
A simulator of a human research community, identified by the lab's GitHub organization as an ICML 2025 project.
An interactive, extensible and controllable framework for building research agents, listed on Jiaxuan You's research page as an EMNLP 2025 project.
Market Position
U Lab positions LLMRouter as a unified, open-source infrastructure layer for LLM routing rather than another single routing policy: it brings disparate router families, benchmark construction, fair quality-cost evaluation and deployment into one workflow. Its stated differentiation is reproducibility and shared infrastructure across routing methods, with the paper reporting a 14.6% relative improvement over the strongest fixed-model baseline in its empirical study.
Leadership
Executive Team
Jiaxuan You
Principal Investigator / Lab Director; Assistant Professor, Siebel School of Computing and Data Science, UIUC
Received Ph.D. and M.S. degrees in Stanford Computer Science under Jure Leskovec; was a core founding member at Kumo AI from 2021 to 2023, where he built a graph-learning predictive system for relational databases; later contributed to NVIDIA's LLM pretraining and posttraining pipeline as a Senior Research Scientist.
Business Model
Revenue Model
U Lab is an academic research lab and its principal software output, LLMRouter, is presented as open source. The reviewed sources describe research code, benchmarks and deployment tooling rather than a commercial monetization model.
Target Markets
- Academic and industrial AI/ML researchers
- Developers building multi-LLM applications
- Teams optimizing response quality and inference cost
- Research-agent and scientific-computing developers
- Users of multimodal, memory-augmented, time-series and personalized AI systems
- Cost-aware selection among multiple LLMs for production requests
- Single-turn, multi-turn and personalized LLM routing
- Memory-augmented reasoning
- Vision and video understanding
- Time-series tasks
- Research-agent and scientific-discovery workflows