Healthcare & Regulated Data

Clinical systems built for the ward, not the demo.

Offline-first. Consent-controlled. Audit-ready.

Deployed

In active clinical use

Multi-centre

Intensive-care study

Offline

First, not as a fallback

A vital-signs monitor and a wall clock in an otherwise empty clinical room

Overview

Clinical data capture and quality systems for environments with unreliable connectivity and strict handling rules — plus the device products built around them. Entries are captured locally, validated at the point of entry, and synced when a connection appears. Consent is a first-class field, not a checkbox bolted on before launch.

The Problem

Ward software fails on the realities of the ward. Connectivity drops mid-round. The tablet is operated one-handed, in gloves. A study needs validation strict enough to survive an ethics committee, and a permission trail that can be revoked field by field. Most systems assume a desk, a network and an unhurried user — none of which exist where the data is actually generated.

Our approach

We build offline-first with conflict-safe sync, so a lost connection is a normal state rather than an error path. Validation runs at the point of entry, where the person who knows the answer is still standing there. Consent and purpose limitation are modelled in the schema on day one — retrofitting them into a live clinical dataset is a rewrite, not a migration.

Deliverables

Offline-first capture with conflict-safe sync
Study-grade validation and audit trail
Consent and purpose-limitation data model
Unit dashboards and period reporting
Gloved, one-handed interface design
Record handling built against DPDP and GDPR

Tech stack

Android / KotlinReact NativeFastAPIPostgreSQLOffline-first syncRow-level isolationAudit loggingDPDP 2023DSGVO / GDPR

In practice

Sync is a data-model decision, not a network feature

Offline-first only holds if the merge rules are agreed before the first tablet ships. INDO-PIX is offline-first with conflict-safe sync, so a record written on a ward with no signal is reconcilable rather than merely stored, and resolution follows the rule the study agreed rather than the order in which tablets reconnected. Those rules are settled with the clinical lead during technical alignment and written into the architecture record, because a study cannot revise its provenance model after collection has started.

Consent that can be answered, field by field

A consent flag is not a permission model. We model purpose limitation in the schema on day one: which field was collected, under which permission, by whom, and what becomes of it when that permission is withdrawn. India-facing platforms are built against the DPDP Act 2023 and DPDP Rules 2025, with full compliance due May 2027; German and EU work is delivered against DSGVO / GDPR obligations, including processor agreements. These are engineering positions taken in live engagements, not certifications, and the proposal states them before it states a price.

How it is verified, and where it fails

The failure mode in clinical capture is not a crash. It is a field quietly absent, or a tablet that stopped syncing and was not missed. So validation runs at the point of entry, where the person who knows the answer is still standing there, and the audit trail is written to be read by an ethics committee rather than by us. The dataset and its history must reconstruct from an export inside one working day.

The device line, and where its boundary sits

This practice includes hardware. VeinMate Assist is a forearm-worn device that images the vein map beneath the skin, selects the insertion point and sets the depth and angle of the needle. It is under active development; features, design and availability are subject to change, and no regulatory clearance or clinical validation is claimed for it. The boundary is the design constraint: the device assists a decision the clinician still makes and an action they still perform. On-device imaging and inference, calibration across every skin tone, a sealed single-use cartridge, under 300 grams.

Where this has run

INDO-PIX

An offline-first Android tablet application for structured intensive-care data collection in a multi-centre clinical study — captured locally, validated at the point of entry, synced when a connection appears.

Deployed and in active clinical use, with conflict-safe sync, a study-grade validation and audit trail, and an interface optimised for gloved, one-handed use.

VeinMate Assist

An AI-guided vein access device worn on the forearm: on-device vein imaging, insertion-point selection, and a single-use mechanical guide that sets depth and angle while the clinician inserts the needle.

Under active development. Under 300 grams with no fixed installation, attach to draw in under 60 seconds. No regulatory clearance or clinical validation is claimed.

ICU Quality Tracker

A quality and outcomes system for intensive care — care-bundle compliance capture, device-associated infection rates, and unit dashboards a consultant can act on during a round.

Proposal issued to the clinical lead, with period reporting and outcome indicators scoped.

Questions

What buyers ask about this specifically.

Yes, and that is the default rather than a degraded mode. INDO-PIX captures and validates locally, then syncs when a connection appears. Where an institution is under a residency mandate, the intelligence layer follows the same rule: if the record cannot leave the country, neither can a prompt containing it, so we serve open-weight models on hardware you control. We have run that pattern with the network interface physically down, not merely firewalled.

Only what we have actually done. We hold no ISO, SOC 2 or HIPAA certification and do not present one. What we bring is a regulatory architecture section in the proposal, and live engagements built against the DPDP Act 2023, the DPDP Rules 2025 and DSGVO / GDPR including processor agreements. Frameworks built against are an engineering posture, not an accreditation. For VeinMate Assist: under active development, with no clearance or clinical validation claimed.

You do. Source code, schema, infrastructure definitions and documentation are yours, on an open-source base carrying no per-seat licence at the platform layer. The test we set ourselves is exportability: if you could not reconstruct your records and their history from an export inside one working day, it is not a platform you own. For a multi-centre study that matters more than usual, because the dataset outlives the software.

Related services

Enterprise Software

Complex systems, reliably delivered.

Microservices. Legacy integration. Long-term ownership.

Product Design

Design that thinks in systems.

UX/UI, design systems, accessibility.

Start a conversation All services