Selected research & engineering.
Projects I use to explore ideas, reproduce methods, build research tooling, and turn model concepts into working systems.
HierarchicalNets
Multi-level hierarchical classification for fine-grained yoga pose recognition.
PoseVLM
Geometry-aware vision-language reasoning for human poses, combining visual evidence with 3D body structure for fine-grained understanding and natural-language feedback.
PhyViT-GAN
Physics-guided visual modeling that combines MobileViT-style representations with generative learning.
CNNVIZ
A compact project for visualizing convolutional representations and making deep-network behavior easier to inspect.
Classical → Modern CV
A hands-on implementation path from classical computer vision through modern deep-learning architectures.
AI Roadmap: Scratch → Frontier
An implementation-driven path for rebuilding core AI foundations and progressing toward current frontier-model concepts.