Multimodal Perception
Fusion models for autonomous driving, indoor localization, and explainable perception systems.
Research Interests
I am interested in building AI methods that connect representation learning with robust engineering deployment, especially in perception, inspection, biomedical signal analysis, large model training, and agent workflows.
Industrial Intelligence
Defect detection, engine room inspection, equipment state monitoring, and vision-based quality control.
Medical AI
EEG seizure detection, graph-based biomedical modeling, survival analysis, and drug-target interaction prediction.
Large Models and Agents
Multimodal large model design and training, RAG systems, private deployment, and AI agent workflows.
Selected Publications
Selected first-author, co-first-author, accepted, and active works from my research record.
LPDD-SSAN: Limited Physics Data Diffusion Guided Selection Sparsity-aware Attention Network for Engine Valve Wear Prediction
Engine valve wear prediction with limited physical data, diffusion guidance, and sparsity-aware attention.
MFICNet: An Multi-modality Fusion Network with Information Compensation for Accurate Indoor Visual Localization
Multimodal information compensation for robust indoor visual localization.
SynerGNN: A Synergistic Graph Neural Network Fusing Database Knowledge and Cohort-Specific Patterns for Cross-Cohort Survival Analysis
Graph learning for cross-cohort biomedical survival analysis using database knowledge and cohort-specific patterns.
Gradient-Guided Parameter Mask for Multi-Scenario Image Restoration Under Adverse Weather
Parameter masking and image restoration for adverse weather scenarios.
AARMC-Net: Axial Adapt Routing Attention Network with Multi-Level Cascaded Neck for Strip Steel Surface Defects
Surface defect detection for strip steel using routing attention and multi-level feature aggregation.
ARD-GNN: Adaptive Recurrent Diffusion Graph Neural Network Leveraging Spatiotemporal Dependencies for Enhanced Seizure Detection
Graph neural network modeling for EEG seizure detection with spatiotemporal dependencies.
Multi-source Heterogeneous Feature Fusion Network for Drug-target Interaction Prediction
Heterogeneous feature fusion for drug-target interaction prediction in biomedical AI.
Other publications and patents
- Interpretable multimodal perception for intelligent driving based on information theory, Science China Information Sciences, 2023, accepted.
- Path planning algorithm for marine engine room inspection based on RSA*-DDWA, RICAI 2024, accepted.
- Practical predefined-time sliding mode control for uncertain marine natural gas engine, CCDC 2024, accepted.
- Patents on graph-convolution-based visual-laser fusion 3D object detection, mechanical defect detection, game-interaction-based explainable perception, and adverse-lighting intelligent driving perception.
Projects
Research and engineering projects across industrial inspection, medical AI, and intelligent connected vehicles.
2024.06 - 2026.12
Fusion Perception Technology for Intelligent Connected Vehicles
Designed multimodal fusion perception system architecture, data specifications, and service interface standards for intelligent connected vehicle applications.
2023.08 - 2024.08
High-precision EEG Seizure Detection
Developed EEG data modeling methods and graph-structured detection networks for clinical seizure prediction and auxiliary diagnosis.
2022.12 - 2024.12
Explainable Machine Learning for Intelligent Perception
Designed explainable perception methods and uncertainty evaluation for unmanned platforms using information-theoretic modeling.
2022.06 - 2024.06
Intelligent Inspection Robot for Ocean-going Vessel Engine Rooms
Contributed path planning, autonomous obstacle avoidance, and industrial defect perception algorithms for engine room inspection robots.
Experience
Applied AI research, model development, and engineering delivery experience.
2024.12 - Present
Founder and Director
Harbin Haishi Yunyun Intelligent Technology Co., Ltd.
Leading AI technology services across private large-model deployment, RAG, agent workflows, computer vision, NLP, time-series forecasting, data governance, MLOps, and cloud-edge-device deployment.
2023.08 - Present
Algorithm Scientist
4Paradigm
Designing deep learning models for biomedical applications including drug-target localization, protein prediction, and molecular biomarker identification.
2022.08 - 2023.08
Algorithm Engineer
Beijing CICV Intelligent Connected Vehicle Technology Research Institute
Worked on multimodal autonomous driving perception, LiDAR-RGB explainable networks, and panoramic multi-task driving perception models.
Education
2025.07 - 2028.06
Fudan University
Ph.D. in Computer Science and Technology, School of Robotics and Advanced Manufacturing.
2022.09 - 2025.06
Harbin Engineering University / Tsinghua University Joint Training
M.Eng. in Electrical Engineering.
2018.09 - 2022.06
Beijing University of Chemical Technology
B.Eng. in Automation, College of Information Science and Technology.
Awards
- National Third Prize, Challenge Cup
- Bronze Award, 9th Internet+ Innovation and Entrepreneurship Competition
- National Third Prize, GigaDevice Cup
- Silver Award, May Fourth Cup Innovation and Entrepreneurship Competition
- Silver Award, 10th Internet+ Innovation and Entrepreneurship Competition
- Third Prize, Carbon Neutrality Innovation and Creativity Competition
- First-Class Scholarship
- National Encouragement Scholarship