I am an undergraduate in Software Engineering at Wuhan University. My current work focuses on computer vision, LLM reasoning, and multi-agent systems, with ongoing research in Universal ReID, scientific graph extraction, and appearance-variant ReID survey writing.

• Status: ECCV 2026 Submission
• Problem: How to inject 3D geometric priors into Universal ReID without being overly sensitive to noisy geometry estimates.
• What I did: Built a reliability-aware 2D–3D fusion framework around SMPL-based geometric representation.
• Evidence: Multi-benchmark training, comparison, and ablation analysis across diverse ReID settings.

• Status: Ongoing, planned submission to SDM
• Problem: How to make scientific graph extraction deployable without depending on frontier LLM inference at scale.
• What I did: Constructed ARCHE-derived supervision signals and a small-model extraction framework.
• Progress: Supervision construction, evaluation workflow validation, and 0.8B-scale training pipeline setup have been completed.

• Status: Ongoing, TPAMI survey preparation
• Focus: ReID under clothing, illumination, and pose changes.
• Current progress: Task taxonomy, evaluation limitations, and metric design space are being organized.

• Computer Vision: Universal ReID, SMPL-based geometry modeling, robustness under appearance and modality changes
• LLM Reasoning: Structured graph extraction, supervision design, deployable small-model reasoning pipelines
• Multi-Agent Systems: Routing, validation, observability, and collaboration workflows under uncertain user intent

• Built an A2A-based multi-agent workflow on the Holos platform.
• Focused on routing, validation, observability, and uncertain-intent handling.
• Demonstrates system-level reasoning workflow design rather than prompt-only integration.

• Worked on software architecture, communication logic, and embedded driver/system integration for a modular robot platform.
• Project won National First Prize (Top 6) in the Huawei Cup IoT Design Competition.
• Demonstrates hardware-software co-design and engineering reliability.

• Participated in 3D interaction module design and prototyping for an AI-assisted digital heritage system.
• Focused on immersive visual presentation and AI-generated content integration.
• Demonstrates cross-modal system design and interactive content delivery.

• Computer Vision
• LLM Reasoning
• Multi-Agent Systems

• Person ReID, SMPL-based human modeling, 3D Gaussian Splatting
• Multi-agent system design, routing, validation, observability
• Structured graph extraction, evaluation workflow design, ablation analysis

• Python, C/C++, Java
• Git, MATLAB
• RISC-V / ESP32 embedded platform development
• Mainstream LLM API-based application development

• Wuhan University, B.Eng. in Software Engineering (Expected 2027)
• GPA: 3.74 / 4.0

• Email: 2023302143001@whu.edu.cn
• GitHub: https://github.com/BohanSu
• Resume: 苏渤涵简历.pdf