Reinventing Bahmni: Modernizing the Stack, the Delivery, and the Product of a Decade-Old Open-Source EMR with Open Standards and AI

A few years ago at IndiaFOSS, the Bahmni community reflected on lessons from years of building a free and open-source Hospital and Electronic Medical Record system for under-resourced health systems. Today Bahmni runs in 500+ sites across 51 countries, and is reinventing itself on three fronts. This is a candid account of what worked and what didn't.

• Modernization - Bahmni carries a 12-year-old, million-line codebase, including roughly 164K lines of legacy AngularJS. Rather than a risky big-bang rewrite, we are incrementally migrating to React and FHIR APIs using a "Research, Review, Rebuild" approach (documented with Martin Fowler), with AI-First Software Delivery applied across design, analysis, development, and testing.

• Delivery and community - AI didn't just make us faster; it lowered the barrier to contributing. Through "Micro-Engagements," developers with only a two-week window now make meaningful changes, because AI-assisted onboarding compresses a week of setup into a single day, turning spare capacity into merged fixes and reusable community plugins.

• The product. We are adding optional, standards-first AI: a conversational, voice-driven interface that turns natural language into structured records (an "AI scribe"), and Clinical AI Insights that surface guideline-aligned suggestions for specialties like diabetology and oncology. Crucially, Bahmni ships the rails, not the engine. It is built on FHIR, SNOMED, and ABDM so adopters avoid vendor lock-in and stay within responsible, compliance-aware boundaries.

Whether you build open-source products, modernize legacy systems, or are figuring out how to adopt AI responsibly, you'll leave with reusable patterns and an honest view of the pitfalls.