- Working on LLM Inference (Amazon Rufus): multi prompting, agent platform

AI powered email assistant to help solve email overload. Working on backend machine learning algorithms and mobile/watch apps.

Develop algorithmic software to make sense of social network data (Twitter, Facebook) using text analysis, NLP and machine learning techniques.- Architected and wrote Relation Extraction engine- Architected and implemented Activity Detection engine- Architected and developed Entity Detection engine- Developed short text classification engine- Now known as InboxVudu

Instant social graph-powered book recommendations. Developed algorithmic software for BookVibe.com book recommendations from Twitter and Facebook stream data. In particular:- Built entity detection and behavior detection for books.- Built user profiling and Twitter influencer detection to improve book recommendations.

- Worked on various topics on detecting and parsing articulated objects (such as people) and recognizing human actions. - Developed effective approximation algorithms in doing inference and learning on highly structured interdependent representations of human body parts (torso, head, arms, legs) to learn to localize body parts.

- Developed visual matching algorithms to better categorize large scale objects in retail stores to prevent ticket switching at self-checkouts. - Collaboratively worked on the project of finding people in surveillance videos based on human attributes.

Taught courses: C/C++ programming, Data Structure, Building Web Applications.