50+ clinics on a HIPAA-compliant telehealth platform

The situation

A US healthcare team came to us with a specific gap: existing telehealth platforms delivered audio and video, but a physician on a call cannot examine a patient. They needed Galen, a platform where a clinic nurse operates the diagnostic hardware on-site while a remote doctor conducts the full consultation, directs device placement, records vital signs, and issues signed clinical documents.

The nurse streams four simultaneous feeds (main camera, handheld examination camera, otoscope, digital stethoscope audio) via Twilio Video. The doctor sees all feeds in a role-specific view, marks a shared body diagram to tell the nurse exactly where to place each device, and at session end signs the prescription and SOAP note through Zoho Sign. Twilio server-composes every session into a single MP4 audit trail for reimbursement.

Fifty-plus clinics onboarded in the first 12 weeks, including rural settings where people previously had to travel for routine checkups. The client reported in-person visits dropped 60%.

What we had to solve

• Four device feeds from one nurse device, with zero setup during the visit

  A nurse in a clinical visit cannot pause to configure a multi-camera streaming setup. The platform detects and starts all four feeds (main camera, examination camera, otoscope, and stethoscope audio) from the nurse's device automatically, each with its own on-off switch visible only to the nurse. The session recording is composed from the doctor's screen at the end of the visit. What the clinic gets: the nurse spends zero time on technical setup during a live patient visit, and every session produces a timestamped record that supports insurance reimbursement.

• Patient-data protection built into every layer, still fast on rural bandwidth

  HIPAA compliance is not a layer added after the platform is built: it is a constraint on every design decision from day one. We enforced patient-data access at two levels. First, the database itself only returns records a role is allowed to see, so a query for one patient can never leak another patient's records. Second, every server function double-checks that the caller owns the appointment before acting. All video runs on a HIPAA-eligible provider with a signed Business Associate Agreement, and every prescription is e-signed through Zoho Sign. What the clinic gets: privacy is enforced structurally rather than by hoping application code behaves, and the whole model still performs on rural low-bandwidth connections.

• Doctor-to-nurse direction without a second system to maintain

  The doctor needs to tell the nurse exactly where to place each device: a mark on a body diagram, not a verbal instruction like "a little lower." Building a dedicated real-time messaging service for this would have added infrastructure, latency, and a new failure point. Instead, the placement instructions, device toggles, and control signals travel through the same video layer that carries the streams, so there is no separate service to maintain or expand.

  Trade-off

  The trade-off: the app itself has to keep the doctor's and nurse's screens in sync. The result: one infrastructure layer instead of two, and no separate messaging service to maintain.

1. Sessions are recorded automatically, audit trail ready for reimbursement

   Twilio publishes the doctor's full screen as a named Video track and server-composes it into a single MP4 at session end: a complete, timestamped record of what the team saw, heard, and decided. The session interface captures SOAP notes during the consultation. At close, the doctor e-signs the prescription and SOAP note through Zoho Sign; Zoho Sign stores signed PDFs, immediately downloadable as the reimbursement record. No separate document workflow. No manually assembled audit trail.

   

2. Nurse operates the peripherals; doctor sees and hears everything, directs placement in real time

   The nurse streams four simultaneous Twilio Video tracks: main camera, handheld examination camera, otoscope, and digital stethoscope audio. Each track has its own toggle in a nurse-only device-status panel; the doctor's view shows all feeds without device controls. When the doctor marks a location on the shared body diagram, a Twilio DataTrack JSON message delivers the placement instruction to the nurse's screen instantly. No verbal reinterpretation, no delay. The doctor directs the examination as precisely as they would in person.

   

3. Every session produces a complete compliance record, BAAs signed from day one

   Two independent layers enforce PHI access. First, Hasura row/column-level permissions per role: a patient's database query cannot return another patient's records before any application code runs. Second, each AWS Lambda function re-verifies that the caller owns the appointment or holds the required role before acting. Video runs on HIPAA-eligible Twilio with a signed BAA. Zoho Sign e-signs every prescription. No compliance control lives only in application logic; the enforcement is structural.

   

4. Patients in rural areas stay connected: poor bandwidth doesn't break the session

   The patients who most need remote care often live in areas with the worst connectivity and the least comfort with technology. They join via a browser link with no login required; there is no app to install, removing the single biggest barrier for rural and low-tech users. We tuned the video and data layers for low-bandwidth connections: adaptive bitrate for video, compressed peripheral data streams, and graceful degradation that maintains session continuity when bandwidth drops. Over 50 clinics in rural and remote settings onboarded in the first 12 weeks without performance issues.

   

How it runs

React

Six different people use Galen (clinic admin, head nurse, nurse, doctor, patient, and our own admin), and each needs a different view of the same live visit. React keeps those views independent so one person's screen never leaks into another's. Its interface library also shipped accessible components out of the box, which mattered for meeting healthcare accessibility standards without building everything from scratch.

Twilio

The consultation room needs four simultaneous video feeds, real-time placement instructions on the body diagram, SMS notifications, and a session recording composed into a single video file. Twilio handles all four feeds from the nurse's device, carries the placement instructions and device toggles between doctor, nurse, and patient as data messages, and composes the recording. One service instead of three means one vendor relationship and one thing to maintain.

AWS

All consultation logic, scheduling, vital signs, prescriptions, notifications, and payments run on serverless AWS infrastructure that meets HIPAA eligibility requirements, scaling from one session to hundreds without anyone provisioning servers. Automated triggers fan out notifications (in-app, email, SMS) and payment-link expiry, so the app never has to poll for updates.

Hasura

Six roles with strictly different patient-data access need enforcement at the database level, not in application code: the database itself refuses queries that a role is not allowed to make. Its event triggers also drive the notification layer, so appointment reminders, payment-link expiry, and status changes fire automatically without a separate scheduler.

PostgreSQL

Session recordings, vital signs, SOAP notes, prescription logs, and audit entries need storage that guarantees every write completes fully. No entry in a HIPAA audit trail can be lost or half-written. PostgreSQL's transactional guarantees make every clinical record complete and defensible for reimbursement documentation and compliance audit.

Stripe

Doctor subscriptions gate platform access; per-booking payments fund clinic operations. Stripe's event-based updates keep subscription status and payment-link expiry in sync automatically: no manual reconciliation, no staff checking spreadsheets.