At AGT Intelligence Pvt. Ltd., I automated geospatial infrastructure analysis with Python computer-vision pipelines, processing aerial imagery and reducing manual analysis time by 40%.
I used Segment Anything Model (SAM) and DINOv2 to extract building footprints and spatial features, achieving 95% precision in automated site selection. I also architected a spatial RAG system with Qdrant for a geospatial AI chatbot.
For my Contextual Memory Framework project, I built a multi-tenant knowledge-retrieval platform with Spring Boot microservices and implemented hybrid retrieval with reranking.
I also developed a PySpark medallion pipeline for Citi’s Data Engineering Coding Workshop, and a dual-stream RGB–thermal model for plant water-stress classification that achieved 91.9% test accuracy.
