Applied AI scientist working on LLM systems, agents, and ML products.

<aside> 🧠

Data Scientist at Optum. My background sits between engineering, data science, and product, so I usually end up working on problems that require both building the system and figuring out what should be built in the first place. I enjoy working on AI features that need to run in production, not just in notebooks.

</aside>

Currently

Data Scientist at Optum, working on healthcare AI systems, agents, and ML workflows.

Previously

Associate Product Manager at Unacademy and Assistant Product Manager at Scenes.

Builder side

I build AI agents, health-tech concepts, product prototypes, and dev-facing experiments.

<aside> ✉️

</aside>

<aside> 🔗

</aside>

<aside> 💻

GitHub

</aside>

Featured Work

🏥 Healthcare Contract Processing Agent System

<aside> 🚀

Enterprise AI system that moved from POC toward a production-facing workflow.

</aside>

Healthcare contract processing is slow, document-heavy, and full of edge cases. I worked on a LangGraph + Google ADK multi-agent system that could process contracts, reason through steps, and reduce manual effort.

My role

Product thinking, architecture, AI workflow design, implementation, and infra migration support.

What I worked on

Designed the agent workflow around document understanding and task decomposition
Worked with LangGraph and Google ADK for multi-agent orchestration
Helped move the ML stack from Azure to GCP and Kubernetes
Thought through reliability, handoffs, failure paths, and operational adoption

Impact

The system grew from POC into a workflow strong enough to replace a 40+ member manual process, with expansion planned across business units.

🤖 Claims AI Agent with Voice and Images

<aside> 🏆

Hackathon-winning claims assistant headed toward production consideration.

</aside>

Built a conversational claims agent to help users and teams navigate claim-related questions and reduce manual resolution effort.

My role

Prototype owner, AI workflow designer, product builder.

What I worked on

Designed the claims conversation flow
Mapped where AI could reduce manual dependency
Worked on prompt design, retrieval flow, and response quality
Focused on operational usefulness instead of a flashy demo

Impact

Won an internal hackathon and moved toward production exploration.

What this proves

I can quickly turn an ambiguous business problem into a usable AI prototype.

💊 RAG Powered Prescription Chatbot

<aside> 🌐

Multilingual healthcare chatbot using retrieval augmented generation.

</aside>

This project focused on answering prescription and healthcare-related queries with grounded context instead of generic LLM replies.

What I worked on

Built a RAG flow for healthcare questions
Added multilingual support for wider accessibility
Designed prompts and retrieval logic for more grounded answers
Treated answer quality and trust as the main product problem

Impact

Won an internal hackathon around claims and healthcare AI.

What this proves

I understand that healthcare AI is not just chat. It needs context, caution, and reliability.

⚙️ CrewAI + ServiceNow Ticket Automation

<aside> 📉

Agentic automation that reduced insurance claims ticket turnaround by 33%.

</aside>

Built an automation layer using CrewAI and ServiceNow MCP to reduce repetitive manual work in claims ticket workflows.

What I worked on

Mapped repetitive ticket resolution steps
Used CrewAI to coordinate agent actions
Integrated with ServiceNow through MCP-style automation
Designed for handoff, traceability, and measurable operational impact

What this proves

I can connect agents to real enterprise tools and make them useful beyond chat.

Builder Projects

Product Management Work

Earlier Product Strategy Work

These are older product pieces. They are not the main thing anymore, but they show the foundation: market research, product thinking, monetization, UX, and strategy.