• Architected a text-to-SQL + RAG pipeline with intelligent routing, dynamically switching between SQL and RAG vector stores based on query intent, serving 150 users across read and summarization workflows, with write-phase agentic execution (natural language → invoice generation) currently in development.

• Eliminated ~20 hours of manual data entry per fortnight for a 2-person ops team by processing 100+ invoices weekly through AI-driven document extraction, replacing typed input with a single-click confirmation flow.

• Built a query complexity scoring layer that routes to Gemini Pro above a 65% complexity threshold and base models below, optimizing inference costs without degrading output quality.

• Designed human-approval gates into the pipeline to validate generated SQL before execution, preventing unsafe queries from reaching the database

• Led product growth engineering initiatives, including onboarding video scripting and nurture funnels, resulting in 40% faster onboarding, 30% reduction in user drop-off, and 35% fewer support inquiries.

• Revamped cost analysis algorithms for recurring expenses, mortgage, utilities, and insurance, using JavaScript and MongoDB, increasing user engagement by 70% through improved financial tracking in weekly digest emails.

• Owned end-to-end development of a Comparable Search feature (React, Node.js, MongoDB), enabling interactive filtering and comparison of insurance providers, contractors, and listings, driving 40% increase in user engagement and becoming one of the platform's most-used tools.

• Integrated TensorLake (Indexify) Document APIs to automate parsing of HTML, PDFs, and structured data, reducing manual data entry by 90% and cutting error rates to near zero across 600+ user profiles.

In my role as a Software Engineer at CrowdDoing, I spearheaded the development of a Wildfire Prevention website using React and Node.js. Leveraging MongoDB Atlas, I optimized data management and outreach efforts. Implementing a CI/CD pipeline, I automated deployment processes for faster website updates.

In my role as a Software Engineer Intern at Cisco, I developed a real-time Agent Tracking UI using TypeScript and Spring Boot, improving developer productivity and reducing manual SQL querying and tracking. I also created a gRPC-based API for agent proxy listing, offering users real-time visibility and control over their proxies while ensuring codebase quality through integration testing.