Technology
Infrastructure your IT team can audit.
Healthcare software earns trust by being unremarkable when it matters. This page is the technical account of how we keep clinical data correct, private, available and portable.
01Architecture
Three layers, shared by every product.
Records, workflows and intelligence are services, not features. Each product calls the same primitives, which is why they cannot disagree about the truth.
01One clinical truth
Record layer
A single patient record written by clinics and readable by patients. Versioned, encrypted, audit-logged and mapped to FHIR R4 resources so it can leave our platform as cleanly as it arrived.
- Longitudinal patient history
- Radiograph and document store
- Consent and access log
- FHIR R4 resource mapping
- Continuous encrypted backup
02The clinic day, in software
Workflow layer
Scheduling, billing, prescriptions, inventory and messaging as shared services rather than features rebuilt per app. Every product calls the same primitives, so a change in the clinic reaches the patient in real time.
- Chair-level scheduling
- Invoicing and reconciliation
- Prescription issuance
- Verified WhatsApp and SMS
- Video consultation infrastructure
03Models that draft, clinicians that decide
Intelligence layer
Retrieval-grounded assistance over the practice’s own record. Every output cites what it read, carries a confidence, and waits for a clinician to accept it before anything is written, sent or billed.
- Ambient clinical drafting
- Treatment-plan suggestions
- Recall and no-show prediction
- Natural-language patient search
- Practice benchmarking
02Engineering budgets
The numbers we hold ourselves to.
Published so you can hold us to them. When we miss a budget we say so in the changelog.
12 ms
Median record write
p50, measured across Indian regions
< 1.8 s
Cold start, clinical screen
Mid-range Android, 3G
99.9%
Availability target
Rolling 90-day, excluding announced maintenance
5 min
Recovery point objective
Continuous encrypted backup
03How it is built
Nine things worth explaining properly.
Not a checklist. Each of these changes what the software can promise a clinic and a patient.
Artificial intelligence
Retrieval-augmented models grounded in the clinic’s own record rather than a general web index. Clinical suggestions are drafts by construction — nothing is written to a chart, sent to a patient or added to a bill without an explicit clinician action.
- Grounded retrieval over practice data
- Human review before any write
- Cited sources and confidence on every output
- Clinician disagreement captured as training signal
Cloud architecture
Serverless compute and managed data services in Indian regions, scaled per tenant. Reads are cached at the edge; writes go through a single authoritative path so two devices in the same clinic never disagree about the truth.
- Multi-region, India-resident data
- Autoscaling functions, no idle servers
- Edge-cached reads, authoritative writes
- Blue-green deploys with instant rollback
Security
Least privilege enforced at the data layer, not just the UI. Access is decided by role claims evaluated on every request, and every read of a patient record is logged with actor, purpose and time.
- AES-256 at rest, TLS 1.3 in transit
- Role-based claims: dentist, assistant, patient, admin
- Row-level authorisation rules
- Immutable audit log of record access
Scalability
Designed for the shape of Indian healthcare: tens of thousands of small tenants rather than a few large ones. Per-tenant isolation means one busy chain cannot slow down a single-chair practice in another state.
- Tenant-isolated data partitions
- Horizontal sharding by practice
- Queue-backed background jobs
- Load-shed gracefully, never lose a write
Microservices
Records, scheduling, billing, messaging, media and intelligence are independently deployable services behind a single gateway. A billing release cannot take down charting, and a model update cannot take down billing.
- Independent deploy per service
- Contract-tested internal APIs
- Idempotent, replayable events
- Circuit breakers on every dependency
APIs & interoperability
Everything the apps do, the API can do. REST endpoints and signed webhooks are versioned with a published deprecation policy, so an integration you build this year still runs next year.
- REST with predictable pagination
- Signed, retried webhooks
- Versioned with 12-month deprecation notice
- OpenAPI schema and typed SDKs
Data privacy
Built to the Digital Personal Data Protection Act, 2023. Patients are data principals with real controls: see who accessed a record, revoke a clinic’s access, export everything, and request erasure where law permits.
- DPDP Act 2023 aligned
- Purpose-limited processing
- Patient-facing access log and revocation
- Data export and erasure workflows
FHIR readiness
Internal models map to HL7 FHIR R4 resources — Patient, Practitioner, Appointment, Encounter, Condition, MedicationRequest, Observation — so that ABDM participation and hospital integrations are a configuration, not a rewrite.
- FHIR R4 resource mapping
- ABDM-compatible identifiers
- Terminology aligned to SNOMED CT and LOINC
- Bulk export in NDJSON
Encryption & key management
Keys are managed in a hardware-backed key store with scheduled rotation and separate keys per environment. Media is served through signed, short-lived URLs; nothing sensitive sits behind a guessable link.
- Hardware-backed key storage
- Scheduled key rotation
- Short-lived signed media URLs
- Secrets never in source control
04Stack
What we run in production.
Chosen for reliability on mid-range devices and weak networks, not for novelty.
Client
Services
Intelligence
Realtime
Operations
Read the API docs, or send us your security questionnaire.
Developers can start with the platform overview and webhook reference. Procurement teams can request our data processing terms, sub-processor list and incident response summary.