University of Minnesota · Industrial and Systems Engineering
Mou GroupLearning, optimization and computation for complex engineering systems Ph.D. openings →

Funded Projects

Our research is supported by the National Science Foundation, Samsung, the University of Minnesota Data Science Initiative, the National AI Research Resource (NAIRR) Pilot and NVIDIA.

Neural Refrigeration Cycle Network enabled Optimal Control of Closed-loop Fluid Infrastructure System

Goal: physics-informed, learning-based controllers for next-generation, energy-efficient control of HVAC systems.

Funded by
From energy-efficient operation to optimal HVAC design: physics-informed learning for large-scale HVAC systems.
From energy-efficient operation to optimal HVAC design: physics-informed learning for large-scale HVAC systems.

Big VISION Initiative

Goal: build a large-scale dataset for vision-based industrial inspection, together with an industry–academia forum and a global platform for sharing data and knowledge on industrial computer vision.

Funded by
  • UMN Data Science Initiative ($200K, no indirect cost; 2025–2028) · DSI news coverage
Big VISION Initiative: what it is, key points and road map
Big VISION at UMN: a large-scale AI dataset, an industry–academia forum and a global platform for vision-based industrial inspection.

Data-Agent: Agentic AI for Autonomous Data Quality Evaluation, Curation and Task-Aware Dataset Design for Vision-Based Industrial Inspection

Industrial inspection is critical to ensuring the safety, reliability, and quality of manufactured products and complex engineering systems, yet advances in AI-enabled inspection are often limited by the availability of high-quality, well-curated data. This NSF-funded project will develop Data-Agent, an agentic AI-powered methodological and software platform for data-centric AI in vision-based industrial inspection. Data-Agent will provide integrated capabilities for automatically evaluating the quality of large and heterogeneous inspection datasets, improving annotations through human feedback, and designing task-aware datasets that are representative and informative for downstream inspection tasks. Building on the VISION Workshop series, an established community for advancing computer vision and AI for industrial inspection, the project will also release VISION V2, together with open-source software, benchmarks, tutorials, and training resources. Through these efforts, the project aims to foster a broader community of researchers, students, and industry practitioners and advance more trustworthy and effective AI systems for advanced manufacturing.

Funded by
  • National Science Foundation, CSSI #2608818 ($600K total, $400K to UMN; 2026–2029)
Team

Shancong (Sean) Mou (PI), with co-PIs Zirui Liu and Yinan Wang

Data-Agent’s three thrusts: VLM-agent dataset screening, foundation-model label curation and task-aware dataset design.
Data-Agent’s three thrusts: VLM-agent dataset screening, foundation-model label curation and task-aware dataset design.

Computing Support

In-kind cloud and GPU computing from the NSF National AI Research Resource (NAIRR) Pilot and the NVIDIA Academic Grant Program, supporting our work on data-centric AI, physics-informed AI and differentiable simulation.

Awards
  • NSF NAIRR Pilot: $238,800 in AWS cloud credits (NAIRR250089; Co-PI) for a data-centric AI framework for defect detection in advanced manufacturing: automated data curation, annotation and design
  • NVIDIA Academic Grant Program:
    • 30K A100 GPU hours (2025–2026) for automatic data quality evaluation and synthetic data generation, supporting the Big VISION Initiative
    • 2 RTX 6000 Max-Q workstation GPUs (2026–2027) for physics-informed AI and differentiable simulation, supporting our HVAC project (a differentiable HVAC system simulator)