QuickBooks AI-Powered Agents Initiative

Delivered breakthrough Agentic AI solutions for payroll automation, aligning cutting-edge AI/ML frameworks with scalable financial operations, leveraging state-of-the-art frameworks (e.g., LangChain and LangGraph) to create multi-agent ecosystems that seamlessly automate complex payroll workflows.

Designed short-term memory (Redis) and long-term memory (persistent datastore) enhancements, enabling agents to recall contextual information and improve personalized interactions for users.

Part of the team which did prompt hardening, agent evals and pen testing providing secured agentic AI experience.

Led the integration of Langfuse observability tools, providing TID-based tracing and unified visibility across agent workflows to improve debugging efficiency, real-time diagnostics, and incident response.

Facilitated organizational adoption of AI-native technologies through knowledge-sharing, model deployment, and reusable frameworks for enhanced operational efficiency.

Spearheaded the development of Anomaly Detection, the in-house classic ML model designed for payroll variance checks, delivering a 57% improvement in annoyance rate detection and achieving a best-in-class error rate of 0.09%.

As a Lead software engineer, I lead a team of four engineers to implement Partner Integration Re-architecture, work included designing a new backend REST solution using AWS Lambda and GraphQL REST service and upgrading the UI framework of entire UI framework based on React.

The project's key highlights included Creating a new set of graphQL compatible REST APIs, a new Lambda application and configuration across 3 ecosystem gateways.

In terms of UI over 800+ files changed, 200,000+ lines of code changed, modernization of 40+ UI elements, and changing three build pipelines.

Project was a significant success as it helped deprecate two major partner workflows, an entire database, and a monolith application on Ruby, while providing 85% increase in performance in partner calls. Helped the team achieve Intuit's target foundation state metrics for our entire React.JS based UI.

In terms of UI we removed deprecated direct and indirect dojo dependencies, migrating from grunt build to yarn build, deprecating & replacing legacy UI Components.We also upgraded unit test framework Intern to Jest using AI LLM scripts.

Mentoring junior engineers and facilitate scrum rituals. Collaborating with cross-functional stakeholders including partners, program managers, and UX designers to develop effective strategies that contributed to the project's success.

The new architecture was scalable & partner agnostics and thus, allowed 3 new HR & Workforce benefits partners to integrate with Intuit Payroll using the same platforms, which was a significant achievement for the team.

As a lead software engineer, I had the privilege of leading a team of three engineers in the seamless migration of 65,000 partner customers to from partner SSO to Intuit's ecosystem sign in process.

Work included implementing code for in sign in persona creation, granting correct access and making sure all authentication and authorization standards are met.

My role was critical in designing solutions, implementing POCs, managing migration experiences, and ensuring release deadlines were met. Mentoring fellow engineers and coordinate with Partner contacts to ensure seamless migration.

65000 Active customers were moved out using the experience from SSO to Intuit Sign in and which provided them better logging experience and a more security to their account, like MFA.

This project was a significant success as it involved removing a primitive IDFed server based SSO integration, which improved the platform's security and reduced partner calls by 150,000 per day.

Decommission the entire IDFed server and related architecture and saving thousands of dollars per year cost for the org.

Completed a number of time sensitive customer-centric projects.

Java web application developer for CCPA (California Consumer Privacy Act) project.

Work included research, analyze 300 tables and schema with legal on user pertinent data. Then developed a streaming pipeline to return 38 meaningful files per customer without 0 load on heap/servers.

ReactJS web developer and technical lead for KBA (Knowledge based authentication) project.Write a new react plugin to replace the old JSP pages, resultant system was 30% faster with a fraud prevention rate increase by 20%.

Designed and developed an async event based data manipulation using JMS ActiveMQ and implement local message recovery systems.

Involved in design and development of an event store mechanism to handle the consistency of a data in a micro-service based architecture.