. Working on LLMs for RAG, grounded QnA, and Prompt Engineering to derive use cases for the product.. Developed models for Document understanding and Information extraction from Visually rich documents ( NER task) and two of the models have been integrated in WatsonX product.Areas of focus: Grounded LLMs | Prompt Engineering | RAG | Incremental Learning | Few shot learning | Graph Neural Networks( GNNs)

• Developed Deep Learning-based models to improve the quality of cardiac images by improving both noise and image texture.• Implemented “Single Image Super-Resolution (SISR) “and “Residual Learning” based Deep Leaning models using Keras and PyTorch to reduce the blooming effect of calcified plaques and stents in the cardiac images.

Developed and deployed Machine Learning based Real-time Indoor Positioning System to localize the Wi-Fi devices in a building with an accuracy range of 3-5 meters. Developed hybrid algorithms to utilize Bluetooth beacons and Wi-Fi data to improve localization performance. Performed intensive study on Wi-Fi scanning and hand-over behavior of mobile devices to increase system’s accuracy.

• Joined the team as a backend developer and transitioned into the roles of Systems admin, Database admin and team lead. • Led the conference application project along with developing and implementing the entire backend architecture which has been used by more than 10 world-wide renowned conferences. Developed the backend for web portal which automated the mobile app. dynamic configuration on the fly, peer-to-peer in-app messaging feature, and content management for the application.• Automated servers routine backup and collaborated with other teams for mobile applications and web portals development. Setup hot standby replicas for redundancy and load balancing for Postgres DB. Implemented the automatic hand-over between active and standby servers in case of failures with 1 sec. response time.• Served as team lead for multiple SDE projects under which I held the responsibilities of designing the architecture, developing backend APIs, deciding the timelines, and carrying out the in-depth unit and integration testing.