
2025, undergrad thesis
Spotting rice leaf disease from photos
mAP@50-95 for YOLOv8, ahead of Faster R-CNN and SSD, across 8,040 annotated images from Vietnamese rice paddies.
AI engineer at GM Solutions in Hà Nội. I teach computers to spot things in photos and video, and build systems that let language models answer from your own documents, with sources.
Grew up in Hạ Long. Also an M.Sc. student in Information Systems at HaUI. Off the clock: lifting heavy, dressing loud.

the serious bit. real numbers, no filler.

2025, undergrad thesis
mAP@50-95 for YOLOv8, ahead of Faster R-CNN and SSD, across 8,040 annotated images from Vietnamese rice paddies.

2024, undergrad thesis
faster inference with YOLOv8 than Faster R-CNN, and 85.71% accuracy against 71.43%, even in low light.
2026, at GM Solutions
Vietnamese document Q&A with citations: vector search fused with a knowledge graph, behind per-document access control. Client: xxxxxxxxx. Details: xxxxxxxxxxxx.
2026 to 2027, master's thesis
The plan: hand and body keypoints, a sequence model, then Vietnamese text-to-speech. No results yet. Ask me again next year.
straight from my thesis figures. models are confident. not always right.




lifting, fits, and a camera roll that's mostly gym mirrors.

hometown. grew up next to all this.
tear one off
Got a project, a paper, or a gym split to argue about? Email me.
hi@nguyenha.online