Sheetal Jantikar

Designed and implemented a metrics aggregator to collect and process real-time playback data from Prime Video players using Kinesis Data Streams, Apache Flink, and AWS S3. Enabled identification of the top 20 most viewed titles across multiple geographies during active playback sessions. Implemented a feature in Prime Video’s live events anomaly detection and ticketing system to prevent duplicate tickets for recurring issues across multiple events. Leveraged ECS Fargate, AWS SQS, and DynamoDB to streamline issue tracking, significantly reducing partner team effort on duplicate analysis and improving overall anomaly detection efficiency. Contributed extensively to the improvement of user experience of Prime Video’s anomaly detection system by adding new features on the UI interface using Typescript.

Experience in design and development of automation testing platform and deployment in Apache Airflow to ensure systematic health checks and software upgrades of the Scout platform. Designed REST APIs to implement new features on the intelligent document processing to interact with the database as well as communicate with other modules, in Django. Led the research on developing a parsing algorithm to extract tables from the vendor invoices using OCR output and computer vision techniques in OpenCV.

Built the deep learning prototype for instrument counting app to count the total instruments in a surgical OR using tensorflow and IOS coreML module. Led project EAGLE for the COVID-19 app to determine if the undertaken COVID test is positive or negative by developing classifier model in Keras with 99% accuracy. Developed end to end pipeline for SpongeCounting App to count the surgical sponges in an image of an OR using deep learning and computer vision.

Responsible for developing and deploying the deep learning based classification and object detection model for detection of drones in real time. Achieved significant performance of 93% on CNN based object detection of drones in cluttered moving environment using YOLO. Worked extensively on image classification networks such as inception-v3, Resnets, Mobilenets and deep learning trackers such as GOTURN.