LabLens - Health Analytics Platform

Inspiration

Lab results are typically delivered as static PDFs with limited context. We wanted to build a structured system that ingests biomarker data, normalizes it against clinical reference ranges, and transforms it into trend-based, analyzable health metrics.

Instead of treating lab reports as isolated snapshots, we modeled them as time-series health data.

What it does

LabLens is a full-stack health analytics platform that:

Stores structured biomarker data across multiple reports (18 core biomarkers)
Normalizes values against defined reference thresholds
Classifies biomarkers into Normal, Borderline, High, and Low states
Calculates report-to-report deltas and trends
Visualizes biomarker progression over time with interactive charts
Generates AI-powered clinician-ready discussion prompts
Provides personalized health insights with profile context

Each biomarker is treated as a typed entity with defined thresholds:

Core Biomarkers (18):

RBC Count, Haemoglobin, Glucose, Creatinine, Urea
Cholesterol, ALT, AST, ALP, Bilirubin
Albumin, GFR, BUN, Sodium, Potassium
Calcium, TSH, FT4

This ensures deterministic and consistent classification across all reports.

How we built it

Frontend

Next.js (App Router) with TypeScript
Component-driven architecture with reusable UI components
TailwindCSS for responsive, accessible design
Accessible semantic HTML (Lighthouse optimized)
Interactive trend visualization with GroupedBarChart component
Real-time profile and report management

Backend / Analysis Layer

Modular API routes for ingestion and analysis
Structured biomarker schema with dual-bound threshold support
Deterministic rule-based classification engine
Delta computation engine for trend analysis:
- Report-to-report comparisons
- Status progression tracking
- Automated insight generation
Persistent profile and report storage layer via localStorage with schema versioning
AI-powered health assistant integration

DevOps

GitHub-integrated CI/CD
Deployed on Vercel
Production optimization via Next.js build pipeline
Schema versioning for seamless data migrations

The system maintains clean separation between:

Presentation layer (React components)
Analysis logic (biomarker classification, delta computation)
Trend computation (time-series analysis)
Data storage (typed TypeScript interfaces)

This architecture enables future ML integration and advanced analytics without rewriting core logic.

Challenges we ran into

Biomarker schema design - Creating a flexible schema that supports both single-bound (high-only, low-only) and dual-bound (normal range) thresholds
Threshold boundary edge cases - Properly handling values exactly at borderline thresholds without false positives
Time-series modeling - Computing meaningful deltas and trends without overengineering
Data persistence - Implementing schema versioning to handle migrations and demo data updates
Accessibility in data-dense UI - Maintaining WCAG compliance while displaying complex biomarker grids
Medical accuracy - Avoiding misleading interpretations while providing actionable insights

Accomplishments we're proud of

Built a fully modular biomarker classification engine supporting 18+ markers
Implemented deterministic status classification with proper dual-bound logic
Designed a reusable trend comparison and delta computation pipeline
Created an interactive multi-marker visualization system
Achieved clean separation of concerns across presentation, logic, and storage layers
Delivered a fully deployed, production-ready build with demo data
Implemented schema versioning for seamless data structure evolution

What we learned

Deterministic health classification requires careful threshold modeling and edge-case handling
Strong TypeScript type definitions prevent subtle logic bugs in medical applications
Separation of concerns dramatically improves maintainability and testability
Accessibility increases trust in data-heavy health applications
Schema versioning is critical for managing evolving data structures
Interactive visualizations make health trends more actionable than static reports