At Thingtrax, I annotate image and video datasets for manufacturing production-line AI systems using Roboflow and CVAT. I also train and validate YOLO object detection models, then review annotations for quality before model-training handoff.
At DenseFusion Pvt. Ltd., I executed annotation and preprocessing pipelines for medical imaging datasets, including mask generation and augmentation. I validated annotations with Dice and IoU metrics to align them with U-Net requirements.
For my MS thesis, I built and annotated a skin-lesion imaging dataset to train a U-Net model, achieving a Dice score of 0.8373. I also curated a labeled skin-cancer classification dataset that achieved 98.56% accuracy.
