15:I[981,["658","static/chunks/658-a698693f56d0116a.js","183","static/chunks/183-6beeeb801c274aaa.js","474","static/chunks/app/%5Btype%5D/layout-631c4730b83c8223.js"],"Posts"] 16:T71c,Hi HN. I’m Akash.

I started working on this after a personal experience where a serious illness in my family went undiagnosed until very late, despite years of lab reports, prescriptions, and scans. The data existed, but it was scattered across portals and PDFs, and no one was looking at it together over time. I ended up copying values into spreadsheets just to see if anything stood out.

That experience made something obvious to me: medicine is longitudinal, but healthcare software treats every visit as a separate event.

This project is a longitudinal health record that brings medical data from different sources into a single, time-ordered history. Patients can upload lab reports, prescriptions, and scans from any provider, and the system organizes them so trends and changes across time are easier to review. On the clinical side, it works as a lightweight record system for outpatient and diagnostic workflows, focused on reviewing history rather than billing or complex hospital processes.

What’s different from traditional EHRs is the emphasis on continuity and portability. The record is built around the patient’s timeline, not a single hospital or visit. Under the hood, we normalize messy real-world inputs and structure them so labs, medications, and diagnoses can be viewed together across time.

This is live today and used by ~25k patients and a few hospitals. We’re still learning where longitudinal views help most in practice, and where they don’t.

You can try it here: https://myaether.live

(There’s a demo flow for uploading records and viewing timelines.)

Happy to answer questions - especially about data normalization, privacy tradeoffs, or why this hasn’t worked well in healthcare before.

How to Fool a Neural Network