Worked on the Computer Vision team in the Ground Truth org

● Designed and implemented an object tracking API using PyTorch. Collaborated with platform, front end and design teams to integrate the API into a product that can be used by annotators. The API helps in reducing annotation times significantly for video sequences.● Implemented a dockerized API that does inference at test time on cloud GPUs. Setup the infrastructure to serve multiple models simultaneously using Tensorflow Serving.● Designed and implemented a service that compares human annotation to model annotation and flags inconsistencies. This helps QA to look at the most egregious errors and understand systematic errors.● Worked with the product team at Amazon to setup Mighty Blur(anonymization service) on Sagemaker marketplace.● Designed and implemented a pipeline for training object detection models using Tensorflow. The pipeline is dockerized and is run on cloud GPUs using Kubernetes.

●Trained a ResNet model for face recognition using Caffe. The model achieved an accuracy of 98.97% on the LFW dataset.●Designed and implemented dockerized micro services to do face detection, face recognition and clustering.

Automatic identification of the type of snowflake in images under Prof Tim Garrett.

Course : ECE 2210, Electrical Engineering for Non-majorsConducting lab sessions.Grading assignments.Proctoring exams