At Corazor Technology Private Limited, I built a multi-modal road-monitoring suite with models for pothole detection, waterlogging segmentation, and crack segmentation. I also engineered a physics-informed model to predict the International Roughness Index.
I architected a Safety Scoring engine combining visual hazard outputs and IRI predictions, with an anomaly classifier to flag non-standard pavement. I deployed a FastAPI and Docker backend serving a five-model inference pipeline.
For SatQuery, I designed a multi-agent satellite visual question-answering system that routes SAR, multispectral, and RGB inputs through vision towers to a LoRA fine-tuned LLM. I added tool orchestration and sync, asynchronous polling, and SSE streaming endpoints for single-GPU inference.
My other projects include traffic-violation detection and evidence analytics, multimodal price prediction, poverty prediction, and cuneiform translation using RAG and LoRA. I’m studying Artificial Intelligence and Machine Learning at University School of Automation and Robotics, GGSIPU.
