Open Research Initiative

AGI for Engineering

A4E is an open research initiative focused on AI-assisted engineering design, initiated and led by Bin Hu from ECE at UIUC. We focus on developing benchmarks, agents, and tools that advance engineering intelligence across multiple domains such as control systems, integrated circuits, computer architectures, structural engineering, mechanical/aerospace systems, signal processing, and others.

A4E
Benchmarks
Agents
Multi-Domain

Advancing AI for Engineering

Developing tools to evaluate and improve AI for engineering.

Simulation & Real-World Data

AI-generated designs undergo rigorous testing through executable simulations and validation against physical data, moving beyond textbook knowledge to genuine engineering capability.

🔬

Multi-Domain Coverage

Spanning control systems, analog integrated circuits, digital hardware, mechanical systems, robotics, signal processing, structural design, and more.

🌐

Open & Reproducible

All benchmarks, code, and datasets are open-source, enabling the research community to build upon and extend this work.

Research Themes

We work on four main areas.

📊

Benchmarking & Evaluation

Developing rigorous benchmarks that measure AI capabilities in practical engineering design tasks. Moving beyond textbook levels to open-ended design problems with simulation-based verification and physical real data.

🤖

Agentic Design

Building LLM-based agents that can autonomously navigate the engineering design process, from requirement analysis through iterative refinement to final validation.

🔧

Tool Integration & Safety Verification

Connecting LLMs with engineering software ecosystems such as MATLAB, SPICE, CAD tools, and domain-specific simulators, while ensuring correctness and safety through formal verification and robust validation pipelines.

🎯

Finetuning & Specialized Models

Finetuning and developing specialized language models tailored for engineering domains, enabling efficient on-premise deployment while preserving proprietary design knowledge.

Featured Projects

Representative Projects pushing toward engineering AGI.

// 01

EngDesign NeurIPS 2025

A community-driven, open-source benchmark evaluating LLM capabilities in real-world engineering design tasks. 101 tasks across 9 domains with simulation-based evaluation.

Benchmark Multi-Domain

Core Contributors

The team behind A4E.

BH

Bin Hu

Initiative Lead
Assistant Professor, ECE @ UIUC
XG

Xingang Guo

Main Student Contributor
PhD Candidate @ UIUC

Plus 50+ collaborators across multiple institutions. See individual project pages for full author lists.

Latest News

Dec 2025

AGI4Engineering initiative launched Announcement

Launch of the AGI4Engineering open research initiative website.

Sep 2025

EngDesign accepted at NeurIPS 2025 Publication

EngDesign benchmark officially accepted to the NeurIPS 2025 Datasets & Benchmarks Track.