Vol. I — Infrastructure
Est. 2015 · Montgomery St.
Canvas Medical
The medical record, rebuilt as a platform you can program - and an AI that documents while you talk.
The medical record, rebuilt as a platform you can program - and an AI that documents while you talk.
A clinician sits across from a patient and simply talks. No frantic typing, no eyes locked on a keyboard, no after-hours "pajama time" spent finishing notes. As they speak, the chart fills itself - history pulled forward, orders drafted, a claim quietly assembling in the background. The record is listening, and for once it is helping. That room is the entire argument for Canvas Medical.
For thirty years the electronic medical record has been the thing clinicians complain about most and change least. It was built for billing departments, not for the people holding the stethoscope. Canvas Medical started from the opposite end: what if the record were a platform - open, programmable, and willing to bend to how care is actually delivered? Not a wall. An API.
Andrew Hines was a data scientist and software engineer - the kind of person who looks at a frustrating system and sees a fixable one. The frustration, in this case, belonged to his wife, a family nurse practitioner who came home worn down not by patients but by the software she had to fight to chart them. He had statistics degrees from San Francisco State and a master's in management science and engineering from Stanford. He also had a problem worth a decade.
So in 2015 he founded Canvas Medical with a contrarian thesis: the EMR's biggest flaw wasn't a missing feature, it was that it could not be extended. Every clinic is a little different, every care model has its own quirks, and yet the dominant systems forced everyone into the same rigid template. Canvas would treat the chart like modern software - something developers could read from, write to, and build on.
The EMR company accelerating everyday medicine.— Canvas Medical's own description of the job
The early years were the unglamorous work of becoming legitimate infrastructure. A seed round in 2018. The first clinic live that same year. A Series A in 2020 that arrived alongside a payer partnership with Anthem - a signal that a startup's programmable EMR could be taken seriously by the establishment. Then, in 2022, the milestone that matters most in this industry: ONC certification, the regulatory stamp that lets an EMR operate as a real, billable system of record. It came bundled with a $24M Series B led by M13.
Strip away the category jargon and Canvas is three things that rarely live together: a certified electronic medical record, a developer platform, and a roster of AI agents - all sharing the same patient data.
An ONC-certified record with care coordination and management workflows built in - and configurable per specialty rather than forced into one mold.
Read and write to the chart, build plugins, and ship new clinical workflows as code. On Canvas, a new workflow can be a pull request instead of a feature request.
An ambient copilot that listens to the visit, writes the note, and places orders - running on the full patient record with clinical guardrails. The code is open.
A Claim Agent and Parsing Agent automate revenue cycle and data extraction, with end-to-end billing built in so practices document and get paid in one place.
The point of all this is leverage. A weight-loss startup, a behavioral-health group, a primary-care clinic for adults over 65 - each can take the same underlying platform and shape it to their model without rebuilding an EMR from scratch. That is why companies like Patina, UpLift, Circulo, and Vivante run on Canvas instead of writing their own.
In March 2025, Canvas did something unusual in a market built on locked-down data: it released Hyperscribe, its ambient AI copilot, as open source - code and evaluation framework included. Most healthcare AI is a black box you rent. Canvas published the box.
Hyperscribe is not just a transcriber. It chains tasks. The output of one step - say, documenting a referral - can trigger the next: eligibility verification, an insurance claim, a clinical decision prompt. Canvas calls this composable, and it only works because the agent sits inside the record rather than bolted to the side of it.
It runs on the patient's complete medical history and ongoing data feeds, so it documents with context, not guesses. Orders pass through clinical informatics guardrails before they execute. And because it is open, an organization can swap in its own foundation model - Claude, Gemini, or OpenAI - or tune the whole thing for a specialty.
The record updates live during the visit. The clinician can edit at any moment, and Hyperscribe keeps going, weaving its work into a single coalesced note. Transparency, it turns out, can be a feature rather than a liability.
A new precedent for accessibility, transparency and effectiveness in clinical workflow automation.— Canvas, on the Hyperscribe launch
Backers include IA Ventures, Upfront, Inspired Capital and M13 (Series B lead). Figures from public filings and press; bar lengths are illustrative.
Andrew Hines remains the technical conscience of the company - the founder who turned a household frustration into programmable infrastructure. In September 2024, Adam Farren stepped in as CEO after serving as President and COO. Farren arrived having been chief growth officer at two other venture-backed EMR companies, Elation Health and Osmind - which is to say he has spent his career selling software to one of the hardest buyers on earth: healthcare.
The culture reflects the product: engineering-led, remote-friendly, and unusually committed to open standards. Canvas's whole bet is that clinicians and developers should build together rather than file tickets at each other. Open-sourcing Hyperscribe was not a marketing stunt - it was that belief, made literal.
Return to where we started. A clinician, a patient, a conversation. In the old version, the visit ends and the real work begins - hours of catching the chart up to what was just said, the record always a step behind the medicine. In the Canvas version, the record kept pace. The note is written, the orders are placed, the claim is moving, and the clinician got to spend the visit being a clinician.
Canvas Medical hasn't reinvented medicine. It has done something quieter and, for the people in that room, more useful: it made the software disappear into the background where it belongs. An EMR you can program, an AI you can read the source of, and a chart that finally listens. That is the room Canvas is building toward - one visit at a time.
Official channels · press · code