Working from engineering practice toward research questions, method design, experiments, and reproducible results.

• Computer Vision in real scenes: object detection, RGB-D perception, sample quality filtering, visual weighing, and error analysis.
• Robust visual prediction: occlusion, abnormal posture, background noise, sample distribution shift, and unreliable sample screening.
• Knowledge-enhanced AI systems: document parsing, semantic retrieval, RAG, Agent workflows, and domain-oriented question answering.
• AI4Science / AI for Life Science: interested in scientific data analysis, biomedical knowledge systems, and multimodal modeling.

Pig RGB-D Visual Weighing & Posture Quality Assessment
2025.09 - Present Computer Vision RGB-D YOLO
Participated in a non-contact visual weighing pipeline for real pig farm scenarios, covering segmentation/keypoint extraction, RGB-D alignment, trunk feature construction, posture quality grading, and weighted regression.

• Built experiments on 349 pigs / 33,521 samples.
• Current image-level result: MAE 3.019kg, MAPE 4.52%, R² 0.9693.
• Pig-level aggregation reaches about 3.21% average relative error.

Agent CLI & AI Coding Tool
Java Maven SQLite MCP LSP Prompt Engineering
Worked on a Java-based Agent CLI for local repository understanding, task decomposition, tool calling, code modification, and execution feedback.

• Built around ReAct, Plan-and-Execute, Multi-Agent collaboration, codebase RAG, MCP tools, long-context management, and interactive CLI workflows.
• Focused on code chunking, Embedding retrieval, tool schema cleanup, prompt layering, LSP diagnostics, and generation-diagnosis-fix loops.

Enterprise RAG Knowledge Base
Java 17 Spring Boot Elasticsearch Kafka Redis MinIO
Participated in a document knowledge management and intelligent QA system with parsing, semantic chunking, vector indexing, hybrid retrieval, permission filtering, and streaming dialogue.

• Internship: Nanjing University Suzhou High-Tech Institute, backend development and monitoring platform maintenance.
• Internship: Global Education, document parsing, retrieval, RAG QA, and multi-tenant permission filtering.
• Core stack: Python, Java, Git, Linux, PyTorch basics, YOLO, OpenCV, Elasticsearch, Kafka, Redis, Docker.
• Awards: National Second Prize in China Undergraduate Mathematical Contest in Modeling; MCM/ICM Honorable Mention.