Building Inner-net, a personality capture engine that syncs with a user and adapts to their digital footprint, for true dynamic personalization in AI products. This works like a user personality plugin to any LLM, any AI product, recommendation engine. The core of it is a memetic knowledge graph engine. Journey so far: Theme of my journey has always been centered on how to use an individual's data efficiently, safely but decisively. The data demands to be processed and can benefit the individual and the collective. Built Fluid, a privacy preserving machine learning platform, for multi party data science collaboration between enterprises working with sensitive data. This enabled business cases that were not possible before, metric improvement is a secondary thought. Used multiple technologies, Federated Learning, Confidential Computing, Multi Party Computation, Differential Privacy. Had to pivot away because it takes too long to sell where privacy and confidentiality matters.Also built Color, a chat based shopping concierge and a recommender engine, based on digital footprint, and evolving conversation memory. Pivoted away because the core memetic personalization engine can be put to better use than just shopping. Raised 2M from Accel Partners.

Researched strategies for dynamic pricing in mobile games for fair and equitable rewards for game items.

Worked on developing a question answering NLP engine, for employee training and knowledge transfer. Trained BiDaff.

Worked on developing retail AI systems for large American retailers. Mostly backend and vision based sorting, cataloguing and shelving of inventory. And also a front facing customer app vision model to recognize the right tools to be bought for the customer's broken machine parts and repair requirements.