Inspiration
In Chile’s public primary care centers, routine blood tests expire after 12 months. Even when initial results already show dangerous glucose, cholesterol, or liver enzyme levels. Patients can wait an entire year for a retest, while healthcare teams spend hours manually reviewing thousands of lab reports. This delays treatment, worsens health outcomes, increases costs, and keeps clinicians tied to administrative work instead of patient care.
What it does
LabSense is a web application that automatically processes and prioritizes laboratory results from Spanish-language PDFs.
- Reads real Chilean lab formats: detects patient info, RUT IDs,
[ * ]abnormal markers, and column structures. - Flags critical cases: calculates severity using Chilean medical standards and lab-specific reference ranges.
- Prioritizes care: dashboard lists patients by urgency with search and filtering tools.
- Tracks history: monitors health marker trends over time.
- Built for compliance: secure authentication, audit trails, and a fully Spanish interface.
In testing, LabSense correctly identified a 73-year-old patient with severe diabetes, poor glucose control, and hypothyroidism. Automatically marked as HIGH priority for immediate care.
Challenges we ran into
One of the biggest challenges in AI-assisted coding is inconsistent UI generation when requirements are not crystal clear. Without a defined design system, AI models tend to “guess” visual details, leading to mismatched components, broken layouts, and extra rework.
Accomplishments that we're proud of
- Went from problem statement to 14 validated requirements and a technical architecture in under 24 hours using Kiro’s spec-to-code workflow.
- Analyzed real Chilean lab PDFs before coding, discovering exact parsing patterns (
[ * ]abnormal markers, Spanish terminology, 5-column structures). - Built an implementation-ready spec including TypeScript interfaces, SQL schemas, and parsing logic. All validated with actual patient data.
- Established a healthcare-grade, accessibility-compliant UI design system on day one, enabling Kiro to produce consistent, high-quality output.
What we learned
- Spec-first AI development eliminates scope creep and ensures compliance-critical requirements are not missed.
- Design systems are not just for humans, they dramatically improve AI coding output by removing ambiguity in UI generation.
- Real-world validation early in the process, catches edge cases that would otherwise surface late in development, saving time and avoiding rework.
What's next for LabSense
We plan to integrate AI-powered PDF parsing, predictive analytics for patient risk, and clinical decision support tools to further reduce delays in life-saving care.
Built With
- google-material-symbols
- google-oauth
- javascript
- kiro
- next-intl
- next.js
- next.js-api-routes
- nextauth.js
- radix-ui-themes
- react
- sql
- supabase
- typescript
- vercel
- yarn
