Misha Rana

Worked on a system to detect and count the number of persons for real-time surveillance using object detection and instance segmentation. The solution is currently being used by the Indian Navy. Developed a system using OCR and python to extract keywords, tables and metrics from financial documents and presented them in a structured format. Deployed the system using AWS Lambda and AWS SQS. Created a solution to redact sensitive information from images using OCR and image segmentation. Created a system for signature extraction and verification using image segmentation and Siamese network and deployed it on the clients GCP environment. Built a system to automate the process of NEFT verification resulting in the reduction of manual intervention by upto 85%. Interacted with the clients to understand the business problem and detail their data science requirements; Led a team of 4 in the process of building end-to-end solutions for clients.

Developed a real-time pipeline of road segmentation with camera for the autonomous vehicle using Deep Learning, which involved creating custom datasets and fine tuning them. Developed a novel deep learning architecture for road segmentation on CAIR outdoor dataset.