Debaditya Roy

Building an end to end intent prediction algorithm, that gathers location data from mobile users having the SDK of predict.io, and applying deep learning to predict the intent (purpose) of visits of those users. The algorithm can tell if the intent of a particular visit was home, work, shopping. Integrating and deploying intent prediction algorithm to AWS Sagemaker platform for training and providing live predictions on the fly.

Worked on the cookie sync algorithm that is used for targeting advertisements to the users. The goal was to reduced the overall complexity of the cookie sync algorithm. Implemented an efficient version of the algorithm during the tenure. Benchmarked KV stores Aerospike, Cassandra using comparable data loads from the company to figure out which database is more suited for the backend of the cookie-sync algorithm.

Responsible for building an application to do OCR on a video stream that would be demonstrated in the edge computing framework created with raspberry pi and Pine A 64 devices. For creating the OCR algorithm we have used Apache Flink as the streaming platform ,OpenCV library for image processing and Java as our language of choice. The application was optimized to run on a resource constrained fog computing environment provided by the raspberry Pi’s.

Worked on integrating third party based Java system by writing custom Java based software and SAP PI 7.0 for a major retail chain in India. The technical system landscape was complex as it had a chain of legacy systems, Oracle suite, SAP suite and proprietary software. Handled a major technical integration by gathering requirements and building custom applications using Java single handedly for a major retail chain in India. Among other projects in the banking and pharmaceutical industry

Built a Java application along with a team of three other interns for generating a COBOL template for mainframe developers. Parsing of hugely unstructured COBOL code obtained from the company was one of the major challenges we overcome during the 6 month internship tenure.