Chinthala Sriteja

– Built and deployed a YOLO-based Computer Vision pipeline for Road Safety Marker detection from indoor top-down imagery, reducing manual annotation effort by about 60% and enabling automated HD map extraction. – Developed a LiDAR point-cloud obstacle detection workflow (pillars, ramps, barriers) using clustering and spatial analytics, replacing a fully manual process across 10+ indoor venues. – Designed graph-based indoor routing by extracting hallway skeletons with NetworkX, enabling scalable turn-by-turn navigation with sub-second query response. – Created high-resolution 2D top-down map generation from 3D point clouds, improving indoor parking visualization for satellite and 2D map viewers. – Productionized Computer Vision models with FastAPI, Docker, and Kafka streaming, partnering with cross-functional mapping and product teams to improve downstream map quality and reliability. – Tech: Python, PyTorch, YOLO, OpenCV, Open3D, NumPy, Shapely, NetworkX, Kafka, Docker, AWS, LiDAR

– Built an automated cheque-processing pipeline using YOLOv5 and Amazon Textract, extracting key fields (account number, amount, date) with 95%+ accuracy on real documents. – Implemented Natural Language Processing (NLP) post-processing and regex to improve extraction consistency across varying cheque templates. – Exposed inference services via FastAPI REST APIs for integration into internal banking document workflows. – Tech: Python, PyTorch, FastAPI, OpenCV, AWS Textract, NLP