Working on End2end Compliance Automation by building a multi-agent compliance investigator for Compliance

investigation for different kinds of Fraud patterns.

Worked on the evaluation, testing for this multi-agent system

Powering Coinbase Customer support chat experience with GenAI to automate customer support scenarios

Worked on:

1. Agentic AI with Langgraph

2. Ranking Models

3. Response Relevance & Styling

4. Guardrails

Worked in the Bing Defensive Team, with the aim of truly capturing extent of Problematic, Offensive and Hateful content of various kinds for all of our Bing Search experiences such as WebResults, ImageResults, AutoSuggest and other generated search results and suggestions. Leading the project as a Tech Lead in a team of 4 people with a keen focus on Metric Quality, and Metric Automation

Worked in the Bing Defensive Ranker team where I built document/webpage level DL transformer based models to identify problematic content of various kinds such as hate, self-harm, misinformation. Worked on optimizing and hosting these models to infer at scale of 400B Web & Image Index. Worked on Bing Core Ranking stack to investigate and ship ranker changes to demote these identified problematic URLs for all user queries on both web and image search results.

Worked on creating a Big Data Platform for Security Researchers & Data Scientists to analyse and detect suspicious behaviour

Technology used - Scala, Python, Azure Synapse Analytics