A passionate AI Developer focusing on Multimodal AI, Agentic Workflows, and Computer Vision. I specialize in bridging the gap between mathematical foundations and high-performance engineering deployment.
- 🔬 Focus Areas: Large Language Models (LLM) / RAG / Multi-Instance Learning (MIL) / Transformer Architectures.
- 💻 Tech Stack: Python, C++, PyTorch, SQL.
- 💡 Philosophy: Solving complex, long-tailed distribution and sparse feature problems with elegant mathematical abstractions and robust engineering.
- 🔗 Agentic-Evaluation-System: 基于 RAG 与 Few-shot Prompting 的多模态大模型辅助决策与评分智能体,支持人机协同闭环 (Human-in-the-loop)。
- 🔗 [WSI-MIL-Attention]: 基于 ResNet-18 与 Attention-based Pooling 的大规模弱监督全切片图像病灶挖掘模型。
- 🔗 [UAV-Forest-Segmentation]: 针对长尾分布与小目标稀疏特征的遥感语义分割,引入 Focal Loss + Dice Loss 与 Swin-Unet 架构。[cite: 51, 56, 58]
- Email: auguszp@163.com