Yijiashun Qi
Biography
Yijiashun Qi is a Masters student in Computer Science at the University of Illinois Urbana-Champaign, with a B.S. from the University of Michigan (Highest Distinction). His research focuses on long-context reasoning for Large Language Models, robust Graph Neural Networks, and multimodal learning, with applications in 3D vision, bioacoustics, and battery prognostics. He has published in peer-reviewed venues with 100+ citations, is an active IEEE member, and frequently reviews for hackathons and conferences. He is skilled in Python (PyTorch, JAX), CUDA, OpenCV, and GNN/LLM toolchains.
Research Interest
Long-context reasoning for Large Language Models (LLMs), Robust Graph Neural Networks (GNNs) addressing class imbalance, multimodal learning, 3D vision and Neural Radiance Fields (NeRF) for industrial defect detection, bioacoustics for marine mammal classification, and machine learning for battery State-of-Health (SoH) prognostics.
Abstract
Machine Learning and Computer Vision for Sustainable Resource Management: From Marine Conservation to Industrial Defect Detection
AI and CV can measurably improve how we manage sustainable resource management. This talk covers three researches that I did:
1. Marine biodiversity: LLM-assisted labeling plus multimodal images improved marine-mammal species ID and sped up survey workflows for conservation teams.
2. Manufacturing quality: NeRF-based 3D inspection detects hard-to-see defects on complex parts, reducing scrap and rework.
3. Battery health: time-series models forecast State-of-Health to extend cycle life and plan maintenance.
Across these domains well share what actually worked: robust GNNs that handle class imbalance and label noise; attention-augmented U-Nets for multi-scale ecological segmentation; and hybrid optimization that blends learned models with existing management rules. Well discuss accuracy and latency numbers, failure modes, and how these tools fit into real decision-making for sustainable resource management.
