Wilbert Andrew Yonathan

I work on computer vision, efficient deep learning and RAG applications. My projects include lightweight YOLO-based detectors for paddy disease (PaddyLiteX) and marine organisms (MarineLite), visual anomaly detection with frozen DINOv2 features and cited RAG explanations built on BGE embeddings, Qwen, LangChain and LangGraph (DefectRAG), and reinforcement learning agents trained with PPO in Unity ML-Agents (dPickleball). I build and evaluate models in PyTorch, OpenCV and YOLO, then deploy them as demos with FastAPI, Gradio and Docker. I'm currently exploring multimodal RAG and AI agents.