Home on Shreekar Gade

Building an AI Pipeline for Commercial Insurance

Tue, 21 Apr 2026 00:00:00 +0000

A deep dive into building an AI system to extract and understand commercial insurance policies as knowledge graphs, covering the domain constraints, pipeline architecture, and lessons learned.

About

Mon, 01 Jan 0001 00:00:00 +0000

👋🏻 Hi, I’m Shreekar.

I’m a software engineer based in Pune, building distributed systems and AI systems that are reliable, resilient, and production-ready.

I’m particularly interested in workflow + agents orchestration (Temporal), event-driven architectures, and designing systems that don’t break under real-world constraints.

Work #

I’ve worked across both 0 to 1 builds and end-to-end production systems.

At Atlan, I worked on distributed systems powering a Microsoft Teams Marketplace integration - building microservices, handling cross-cloud infra (AWS + Azure), and designing event-driven systems using Kafka. I also spent time making systems more reliable (and less memory-hungry).

Projects

Mon, 01 Jan 0001 00:00:00 +0000

A selection of things I’ve built.

insura-ai — Insura is an AI-powered workspace and assistant designed specifically for insurance operations. Python 2026
ai-interviewer — This application enables user to create a personalized or customizable AI Interviews. TypeScript 2025
ai-workflow-automation — Workflow Automation enables seamless automation of your workflows. TypeScript 2025
queryflow — QueryFlow is an application that simulates running SQL queries seamlessly. TypeScript 2025

Web #

sticky-notes — An application for creating, editing, and deleting sticky notes with public/private visibility. TypeScript 2025

Work Experience

Mon, 01 Jan 0001 00:00:00 +0000

Atlan #

June 2025 - Dec 2025

Architected and launched an official Microsoft Teams Marketplace App replacing a legacy custom integration. Designed a production-ready microservice handling secure cross-cloud communication between Azure and AWS via site-to-site VPN, with distributed transaction coordination using Kafka and the Saga pattern.

Impact: Reduced enterprise customer setup time by ~95% (days → minutes), migrated 25+ enterprise customers with 200+ app installation, debugged and resolved critical OOM issue reducing peak memory from ~750MB to ~30MB.