At IIT Guwahati, I’m architecting a plastic waste segregation pipeline for real-time classification. I implemented supervised contrastive learning for feature extraction and image classification on the WaDaBa dataset.
For the ISRO GeoNLI project, I engineered a vision-language pipeline combining LoRA-tuned YOLOv11 and Qwen2-VL to process satellite imagery. I also developed a spatial reasoning system for geospatial question answering and oriented bounding box grounding.
On Baby Dragon Hatchling, I engineered a post-Transformer architecture using reaction-diffusion dynamics and Hebbian plasticity. I also developed a workbench for visualizing neuron firing and sparsity.
I’ve worked on computer vision projects involving 3D point cloud reconstruction and depth estimation, and built an LLM-driven RPG engine with contextual memory. My technical toolkit includes Python, PyTorch, TensorFlow, and OpenCV.
