Research & projects
Research in AIOps, multimodal diagnosis and LLM/agent applications, alongside practical AI engineering. Exploratory work and completed development are identified separately.
Industry application
Hubei Port Unified External Service Platform
LLM/agent workflows for email classification, structured extraction, knowledge mapping, quality checks and EDI conversion, with human review and automated testing.
Project details →Research and infrastructure
Multimodal root cause analysis for microservices
Investigating joint diagnosis using logs, metrics, and traces, with a focus on cross-system generalization and fine-grained root cause localization. Reviewing approaches to multimodal fusion and fault propagation modeling.
LLM/agent-assisted log analysis
Studying LLMs and agents for log parsing, anomaly understanding, and root cause reasoning. Exploring multi-step reasoning and tool collaboration, with attention to diagnostic evidence and interpretability.
Industrial log analysis and intelligent diagnosis
Investigating log parsing, anomaly detection, and fault localization for industrial metering automation systems, aiming to identify diagnosis approaches suited to real operational needs.
Research experimentation and model evaluation
Building a reproducible experimentation workspace for configuration management, run tracking, metric analysis, and result comparison to support algorithm research and model evaluation.
Previous development
Smart seat monitoring system
Pressure sensing, MQTT communication, model inference and a web dashboard for posture monitoring and multi-user visualization.
Project details →Raspberry Pi-based space mining robot system
Developed motion control, path planning and obstacle avoidance, and implemented vision algorithms for target detection and environmental perception. Contributed to task allocation and collaboration strategies for a dual-robot platform. The project received a National First Prize at the 27th China Robot and Artificial Intelligence Competition.
OCR- and AI-based personnel archive retrieval
Designed a multilingual OCR model combining ResNet, LSTM, and attention. Developed retrieval, structured storage, and access control modules using inverted indexing, fuzzy matching, and MySQL. The project was recommended at the university level.