If you've followed this blog for a while, you've mostly seen posts about testing methodologies, CI/CD pipelines, and the occasional dive into machine learning in .NET. This post is a bit different. I want to introduce something I've been building for the past couple of years alongside my day job: Alehra, a practice operating system for health providers, therapists, counselors, and psychologists.
Why not just use an existing EHR?
Most EHR software on the market was built for general medical practices and later adapted for behavioral health. That adaptation shows up in small but constant friction: note templates that don't match how therapy sessions actually unfold, billing workflows built around procedure codes rather than session-based care, and telehealth bolted on as an afterthought rather than built in from the start.
Alehra is built specifically for behavioral health from the ground up, and that shows up in a few concrete ways. Documentation is ambient rather than manual: the system listens during a session and drafts a clinical note using templates a practice can customize to match their own language and structure, instead of forcing clinicians into a rigid form. An AI agent reviews each note for clinical concerns and helps track progress against a client's treatment plan over time, something general-purpose EHRs typically don't attempt at all. Telehealth runs natively rather than through a bolted-on third-party link, and the same ambient documentation works whether a session happens in person or on screen. Billing is unified across payment providers, with funds going directly to the practitioner rather than sitting with an intermediary.
The result is a single system instead of the four or five disconnected tools a lot of practices are stitching together today.
The engineering behind it
For readers who come here for the technical content, here's a bit of what's under the hood. The frontend is React, and the backend is a mix of microservices in Node.js/TypeScript and Python with FastAPI. The ambient documentation feature runs on a mix of large language models. Everything runs on GCP. Cursor with assistance through Claude Code & Codex have been my primary tools for a lot of the day-to-day implementation work, which is its own interesting story about how AI-assisted coding changes what a two-person technical team can ship.
Where things stand
Alehra started as a white-labeled deployment that grew to serve more than 250 providers and 3,000 patients. We're working through APA certification in the US and Alberta, building partnerships with the Canadian Psychological Association and the Psychologists' Association of Alberta, and starting to plan an expansion into behavioral health markets across Africa.
I'll probably write more here about the technical decisions behind Alehra as we go, from real-time telehealth to how we've approached hallucination hardening in the clinical documentation pipeline. For now, if you want to see what we've built, you can find it at alehra.com.