About Harvis Health
An AI-first company that built the hospital record its models run on.
Harvis Health builds the models that document, read and predict inside a hospital — and, because intelligence is only as good as the record underneath it, we built the hospital information system too. One platform, from the consultation to the patient's phone. Founded in Chennai in 2021, headquartered in Jacksonville, Florida, live in 75+ hospitals today.
Our mission, in seven words: optimize, automate and innovate the healthcare ecosystem.
Intelligence at the core, one record around it. Every model reads from the same live chart and writes back to it — the record we built so the AI would have something true to work from.
What we believe
Five convictions about AI in a hospital — and what each one built.
These are not values. They are engineering decisions about where intelligence belongs, and every one of them shows up as something concrete in the product.
-
Intelligence belongs inside the workflow, not beside it.BuiltYakko drafts the note into the EMR while the doctor is still talking
No chat window. A model's output lands where the decision is made — in the note, on the chart, at the top of the worklist — or it is not finished.
-
A model is only as good as the record it reads.BuiltHarvis HIS — 55 modules, one patient identity, no nightly sync
An engine fed by four systems synced overnight is reading yesterday's hospital. So we built the record ourselves: one live chart, one audit trail, nothing exported and re-imported.
-
Show the working.BuiltGooseberry — a claim with zero citations cannot even be stored
A confident answer nobody can check is worse than no answer. Anything the system claims clinically has to quote the exact row of the record it came from, frozen at the moment it read it.
-
A human signs.BuiltFour decision states, two of them requiring a written reason
Nothing auto-applies to a chart. Notes are drafted for signature; suggestions are proposed for a decision, and the decision — accept, reject or override, with a reason — is recorded against a name.
-
Measure the misses.BuiltHarvis AI — 15+ models, each with its own confirm rate and drift status
Every model drifts as the population it serves changes. The only question that matters is whether it admits it — so every model in the fleet is versioned, monitored for drift, and reported with what it got wrong.
Why we exist
Optimize, automate and innovate the healthcare ecosystem.
That sentence is our mission, and every product we ship has to answer to one of its three verbs.
Make every decision in the hospital a better-informed one.
Models that forecast tomorrow's demand, rank the panel by risk, catch revenue leaking before the invoice closes — and are measured on what they get wrong, not only on what they get right.
Take the typing and the checking off the clinician.
A consultation note that drafts itself while the doctor talks. A second read on every case that quotes the chart it came from, so the clinician can overrule it rather than obey it.
One live record, from the ward to the living room.
Intelligence is only as good as the record underneath it, so we built the record too: 55 modules on one audit trail — and the app that carries personal baselines and 140+ biomarkers home with the patient, connected by consent.
VisionThe hospital we are building toward has fewer errors, less time lost to paperwork, and clinicians who spend their day on patients. Everything on this site is a step toward that hospital.
How the platform grew
Built in layers, in the order the intelligence needed them.
Every layer exists because the models needed it. The record came first, because a model that reads from four systems synced overnight is reading yesterday's hospital.
-
2021Founded in Chennai
A doctor, an AI researcher and a product engineer — and one decision: the record does the work.
Three founders -
Layer 1The record
The models needed something true to read. Harvis HIS ships its first twelve modules — registration, OPD, IPD, pharmacy, lab and billing — on one patient identity and one audit trail.
12 modules -
Layer 2Every department
Emergency, OT, ICU, blood bank, CSSD, dialysis, the mortuary register, the biomedical-waste log — plus 17 specialty packs.
55 modules · 770+ screens -
Layer 3The engines
Yakko drafts the note. Gooseberry reads the chart back. BucketAI ranks the panel. Harvis AI runs the fleet — drift-monitored, in production.
15+ models live -
Layer 4Beyond the walls
Harvis Me puts personal baselines and 140+ biomarkers on the patient's phone — connected, by consent, to the hospital record.
The record goes home -
2026Next
The same platform in new markets — Southeast Asia and the Middle East — with the schemes, taxes and languages each one actually runs on.
Expansion
Only the first and last stops carry a date. The four in between are the order things had to be built in — the record first, so the intelligence on top of it had something true to work from.
What we hold to
Seven rules we build by.
Each of these has stopped something we had already built from shipping. That is the only test of a value that matters.
- 01Patient first
When safety trades against convenience, speed or margin, safety wins and the meeting ends.
- 02Clinician empathy
The test is not whether it demos to an administrator. It is whether a nurse can use it in hour fourteen of a shift, in a dark ward, one-handed.
- 03AI with purpose
Intelligence goes where it removes real work or catches real risk. We do not bolt a chat box onto a form and call it innovation.
- 04Show the working
Anything the system claims clinically quotes the record it came from. A confident answer nobody can check is worse than no answer.
- 05Trust and security
Role-based access, complete audit trails, isolated tenancy, encrypted backups. Nothing happens to a record off the books.
- 06Local reality, global standards
Built for the schemes, taxes, registers and languages a hospital actually deals with — on HL7, FHIR and DICOM, so it travels.
- 07Ship fast, stay stable
Hospitals do not close for a release. Small changes, often, with a way back — against a 99.9% uptime target.
Who builds it
Fifty-two people. Eight of them are doctors.
One in six of the company is a clinician — not an advisory board that meets quarterly, but people who sit in the same design reviews as the engineers, write the clinical logic, and test releases on the floor.
Own the model fleet: training, drift monitoring, and honest reporting of what each model gets wrong.
Build and run the platform — 55 modules, the record underneath them, and the infrastructure that has to stay up while a ward is using it.
On the floor through go-live and after it — the first to find out what we designed badly.
Write the clinical logic, review every screen, and reject the ones that would not survive a real shift.
Remote-first, across Jacksonville and Chennai — and deliberately small. A hospital platform is built by keeping the people who know why a screen is shaped the way it is.
Where we are
Two offices, ten hours apart.
Headquartered in Jacksonville, Florida, with engineering and implementation in Chennai. When one office closes for the day, the other has already opened.
Jacksonville, Florida
13241 Bartram Park Blvd, Building 23
Jacksonville, FL 32258, United States
Chennai, India
No. 54, Mount Poonamallee Road, St. Thomas Mount
Chennai 600 016, India
- Chennai · 09:00 – 18:00 IST
- Jacksonville · 09:00 – 18:00 ET, which is 18:30 – 03:30 in Chennai
Shown in Chennai time; the offices are nine and a half to ten and a half hours apart depending on the season. The help desk answers in between on +91 99408 51611.
Whichever side of the door you are on
Run a hospital on it — or just run yourself on it.
Hospitals get a 20-minute demo on their own specialty, in their own language — a live OPD consult, a discharge and a billing cycle, no slide decks. Individuals get early access to Harvis Me, and a straight answer on price before anything is charged.