Building a multi-device AI biometric screening platform for healthcare

PROJECT OVERVIEW

ABOUT THE CLIENT

This case study is based on a real VNEXT project. The client's identity and certain project details have been anonymized or generalized for confidentiality purposes.

Our client is a Japan based healthcare technology company that provides AI powered biometric screening, analyzing pupillary response patterns to help identify signs of fatigue, stress, or physical impairment in individuals such as athletes and other at risk groups. Exact financials are withheld for confidentiality.

THE CHALLENGE

Building a diagnostic tool around pupillary response analysis meant working in a specialized medical and ophthalmological domain, where a misunderstood clinical concept could shape an entire feature the wrong way.

The AI model at the center of the platform was developed by a separate, independent team outside VNEXT, which meant every schedule depended on coordination across two organizations with different working rhythms and priorities, not just on VNEXT's own delivery pace.

As the platform proved itself, the client kept asking for more: an admin console for managing biodata and monitoring AI analysis errors, a dedicated web experience for end users to view their own results, and eventually a desktop application for branch and system admins who needed to keep working even without a stable network connection, syncing data automatically once back online. Each addition had to fit into a system that was already live and already depended on for real screening work.

PROJECT OBJECTIVE

The client wanted to move beyond a single AI powered analysis tool into a complete screening platform: one where administrators could manage biodata and catch analysis errors quickly, end users could view their own results directly, and branch and system admins could keep working reliably even offline, all without ever pausing the system already in daily use to build the next piece.

APPROACH AND METHODOLOGY

Because this platform grew across a specialized medical domain, an external AI team, and several new device targets, VNEXT's approach had to cover more than writing code well. Four practices carried the most weight.

Building fluency in a specialized medical domain

Rather than rely on the client to fully specify every clinical detail, the bridge engineer took on dedicated research into pupillary response analysis as a domain, then scheduled frequent, focused meetings with the client to confirm requirements against that growing understanding, so specifications reflected real clinical meaning rather than a surface level reading of a document.

Coordinating cleanly across two development teams

With the AI model owned by a separate team, VNEXT's own project management defined shared working rules both teams could follow and kept a close eye on cross team commitments, so a slower response from the AI side surfaced early as a schedule risk rather than as a surprise near a release date.

Adopting each new platform deliberately

Extending the system onto Azure infrastructure, and later onto an Electron based desktop client, meant building solid expertise in each platform quickly and on purpose: engineers dedicated study time before writing production code, worked from official documentation and established best practices, and escalated directly to platform level support channels whenever a technical question needed an authoritative answer.

Confirming direction before writing formal documents

Early user interface requirements shifted more than once as the client's own thinking evolved. Instead of rewriting formal design documents every time, the team used quick diagrams to confirm shared understanding first, only committing requirements to formal documentation once the client had settled on a direction, which kept rework contained to sketches rather than finished specifications.

TEAM SCALE

VNEXT matched team size to the nature of each engagement, from a larger group of engineers during the platform's core build out to smaller, focused teams for later feature specific work such as the desktop app and its end user facing web module, all under the same Dedicated Team model VNEXT uses for other long running accounts.

• Dev Lead / Tech Lead: owned frontend architecture and user experience quality, and mentored the team through technical seminars and documentation
• Backend Developer / Solution Architect: designed and managed API versions, backend security, and reusable frameworks shared across the platform's modules
• QC Lead: owned test quality and delivery timelines, and kept team checklists current as new issues surfaced
• Business Analyst: translated business requirements into detailed specifications and user flows, and stayed the team's point of reference for spec questions

THE SOLUTION AND TECHNOLOGY STACK

What began as a single AI powered analysis tool grew, engagement by engagement, into a full screening platform reachable from an admin console, an end user web app, and a desktop application built for environments where the network cannot always be trusted.

What was built

• AI powered biodata analysis engine and admin console: lets administrators register biodata, review AI analysis results, and investigate flagged errors from a single filterable view
• End user facing web application: gives individual users, such as athletes, direct access to their own screening results and history, through a mobile responsive interface built for checking results on the go
• Offline capable desktop application: connects branch and system admins to the platform even without a live network connection, syncing data automatically once connectivity returns
• A separate, multi language web module: extends the platform to a broader tier of end users, supporting both Japanese and English

Each addition was built to plug into the existing platform rather than stand apart from it, so the admin console, the end user web app, and the desktop client all read from and write to the same underlying biodata and analysis records.

Technology Stack

This engagement sits alongside VNEXT's broader AI development practice, which covers building and integrating AI powered features into production software across healthcare, retail, and other industries.

THE RESULTS

What began as a single AI powered analysis tool now runs as a complete screening platform, used every day across admin, end user, and offline environments.

• One tool became four connected products: the platform now spans an AI powered admin console, an end user facing web application, an offline capable desktop app, and a dedicated multi language module, all built on the same underlying data
• Faster error investigation: administrators catch and investigate flagged AI analysis errors from a single filterable view instead of searching across separate records
• Uninterrupted screening, even offline: the desktop application for branch and system admins keeps working without a network connection and syncs automatically the moment connectivity returns
• Three years of continuous expansion: the client returned engagement after engagement to add new capability to the same platform, extending its reach without ever rebuilding the foundation underneath it

For a healthcare technology company whose product depends on accurate, always available screening, that combination, faster investigation, offline reliability, and steady expansion without a rebuild, is what turns a single AI tool into infrastructure the client relies on every day.