A clinical-stage health technology company specializing in concussion management came to Armakuni for natural language search interface for a concussion management platform, enabling clinicians to query patient assessments across symptom and reaction time data.
The challenge
Armakuni proposed a targeted PoV to show that a natural language query layer could be built on top of C3Logix's existing AWS infrastructure in under 5 weeks. The engagement is standalone, scoped to a single affiliate dataset, and structured to prove clinical utility before a full platform rollout.
Scale: 750,000+ unique subjects, 900,000+ assessments, 3,000 users across teams and affiliates.
What we built
Armakuni delivered natural language search interface for a concussion management platform, enabling clinicians to query patient assessments across symptom and reaction time data. The engagement ran as single fixed-scope PoV: Smart Search Capabilities. AWS funded the work at $11,700.
The outcomes
5 measurable outcomes shipped across the engagement. The ones that moved the business the most:
Natural language access to both clinical modules from a single question. Queries spanning Symptom Checklist and Reaction Time data return one consolidated answer, so clinicians no longer consult two sources and synthesize manually.
Interpreted findings rather than raw values. Results arrive as summaries of average symptom severity, reaction time delta from baseline, and group-level trends, so clinicians read a conclusion instead of a data table.
Patient safety preserved through clinical precision. The system is designed to avoid misinterpreting headache severity scores, emotional symptom indicators, and reaction time changes, and every answer is scoped to the correct team, affiliate, and timeframe.
Proof of value delivered in 4 to 5 weeks at a $1,300 net customer cost. Of the $13,000 total engagement value, $11,700 was covered by AWS funding, proving the capability before any commitment to full integration.
Architecture ready to scale to 750,000 subjects without a rebuild. The Fargate backend, OpenSearch vector layer, and query classifier are built at proof-of-value scale but structured to absorb the full clinical dataset in Phase 2.
Built on AWS
The production environment runs on 14 first-party AWS services. Delivered under our AWS AI Services, Data and Analytics competencies.
What's next
Phase 2 roadmap: integration into the Ruby on Rails web portal, extension to additional modules (balance tests, cognitive assessments, visual processing), HIPAA-aligned role-based access controls, audit logging, and query fine-tuning using Amazon services based on real usage data.