prashant singh

I was working as a full time data scientist for vearly stage startup called Dermasreen UG. Research and development of Deep Learning models centered around classifying the skin cancer using RGB images. Achieved weighted accuracy above 85% on our hold out testset and weighted accuracy of 82% on ISIC public testset. Created a complete infrastructure on AmazonWeb services for model training and result visualisation. Experience with performance analysis, optimizations and benchmark evaluations

Developed image augmentation pipelines for a CRNN models and trained models for reading the contents of driver license Trained a SSD based object detection network to detect the license plates of the car. The Accuracy achieved was 94% based on 50% IoU threshold. The detected license plates was then anonymized.

Trained deep learning models for solar panel detection and segmentation using TensorFlow. Written bash script to work with sync in Amazon S3 and management of multiple server on Amazon EC2 using playbook. Involved in labelling features on Autobahn, Solar panels and pools using Qgis. Implement a multiple logging protection on the website account using Django