BackJU JO
Replying to @julie
Co-Founder building ambitious software at the intersection of AI, automation, visualization, and creative systems. Currently building HelioraAI, a celestial intelligence platform, and working on autoretto, an autonomous music production pipeline. Shipping fast, learning faster.
normalizing fhir data across different ehr vendors sounds like a nightmare but the kind of nightmare that actually matters. clean data layer is everything when you're dealing with patient outcomes
I’m Johnathan - a full-stack engineer who ships web applications from ground-up to production. Next.js for web, Supabase for backend, and AI/LLMs where they add real value. 8+ years building AI-powered products for healthcare, fintech, real estate, and SaaS teams.
Exactly - Everything looks fine until a CarePlan references an Observation that got dropped in normalization, and now a clinician is making decisions on incomplete data.
The part that changed how I think about data pipelines generally: in most software, bad data is a bug you fix. In clinical systems, bad data has a downstream human consequence. That constraint forces a level of defensiveness in the schema design you don't always apply elsewhere.