Purdue biomedical engineering, where I published on whether osteoporosis drugs actually make bones stronger. Then fifteen years building AI systems in healthcare — medical imaging, diagnostic tools, ECG analysis for medical devices, health data platforms. I've shipped across the stack, from DICOM processing to LLM-powered clinical reasoning.
Different companies, different technologies, same lesson: the problems that kill healthcare AI aren't technical. They're structural. Bad data, vendor gatekeeping, regulatory promises without enforcement, incentives that don't align with patients. I kept finding the same walls from different angles, and eventually stopped looking for someone else to fix them.
In 2022 I started Phreable to make medical records usable for patients. I bet on the 21st Century Cures Act forcing health systems to open their data. I built the platform, tested it with real records, and hit three walls at once — models that hallucinated on clinical content, data that was unreliable even when accessible, and legislation nobody enforced. The right call was to stop before putting patients in front of something I couldn't trust.
Every wrong assumption pointed to its inverse. You don't need comprehensive medical records; the most important health data is what the patient already knows. You don't start from all the data and compute an answer; you start with the person and build understanding from there. That inversion became ΔN1.
Colorado. Two kids. I build things.
Projects
- ΔN1 — Health intelligence that learns what works for your body
- Zone Pedal — Heart rate training for cyclists
- Phreable — AI health platform, 2022–2024. What it taught became ΔN1.
mark@markkoivuniemi.com