- 🎓 Student at Zhejiang University
- 🤖 Passionate about AI Infra, AI Agents, Multimodal Models
- 🔧 Contributing to open-source projects in the AI ecosystem
- 🏆 Multiple academic competition awards
LiquidIdentification — Bottle Liquid-Level Detection
Built a pipeline for transparent bottle residual liquid identification using YOLO OBB and traditional ML.
- Custom BottleDataset (218 images) with LCDTC integration
- YOLO11m-OBB rotation detection + 247-dimensional interpretable features
- Compared SVM, Random Forest, XGBoost; achieved mAP@0.5 = 0.811
minimind-llava-v — Lightweight Multimodal Model
Extended MiniMind with vision-language capabilities using a LLaVA-style architecture.
- Adapted vision encoder + language model with projection / cross-attention interfaces
- Training script with mixed precision, gradient accumulation, checkpoint recovery
ATTI — Auto AI Questionnaire Evaluator
Edge browser extension for automated AI questionnaire evaluation.
- Supports MBTI / Truity information extraction
- Session persistence, auto-fill, and memory-bank rule maintenance
- 🔬 Multimodal models and vision-language architectures
- 🏗️ AI infrastructure and efficient training/inference
- 🤖 Agent systems, MCP, and tool-use frameworks
"The best way to predict the future is to invent it." — Alan Kay