SmartCarePlus – AI Lab Report Template Digitizer

Convert any paper or legacy lab report into a structured digital template in minutes using Gemini 3 on Vertex AI. Upload PDFs or images, get validated JSON schemas with formulas and reference ranges.

Comment

Inspiration

Most laboratories hesitate to change their software because of one painful reason — report templates. In fact, over 80% of labs still run locally and avoid moving to modern cloud systems primarily due to the complexity of migrating their legacy report formats.
Every lab has unique paper formats built over years, and migrating them into digital systems requires manual JSON editing or developer time. This bottleneck slows down digital adoption in healthcare.

We built SmartCarePlus – AI Lab Report Template Digitizer to remove this barrier completely. Our goal: allow any lab to convert existing paper or legacy report formats into structured digital templates within minutes using Gemini 3.

What it does

SmartCarePlus AI Lab Report Template Digitizer transforms unstructured lab reports into fully structured digital templates.

Users can:

  • Upload PDFs or scanned images of existing lab reports
  • Provide text prompts describing new test formats
  • Receive a validated JSON LabReportTemplate schema
  • Auto-generated variables, units, reference ranges, and formulas
  • Preview the final rendered report instantly
  • Continue chatting to refine templates iteratively
  • Non-technical lab owners can vibe-code complex report templates using natural language without writing a single line of JSON

This turns days of manual configuration into minutes.

How we built it

The system is a containerized Python service integrated into the SmartCarePlus platform.

Pipeline:

  1. Native multimodal understanding — Gemini 3 Flash directly analyzes PDFs and scanned images using its vision capabilities, extracting structure and semantics beyond simple OCR
  2. Gemini 3 Flash on Google Cloud Vertex AI performs:
    • Document structure understanding
    • Parameter extraction
    • Reference range detection
    • Logical formula inference
    • Structured JSON generation
  3. Schema generation – Outputs strict LabReportTemplate JSON
  4. Validation engine ensures formulas reference valid variables
  5. WebSocket interface enables real-time conversational refinement

Using Gemini 3 Flash via Vertex AI provides low-latency, scalable, production-grade inference, enabling real-time interaction suitable for healthcare environments.

Challenges we ran into

  • Handling noisy scans from low-quality lab printers
  • Teaching the model to infer medical calculation formulas reliably
  • Ensuring strict JSON schema compliance every time
  • Building safe validation to prevent incorrect clinical logic

We solved this using structured prompting, response validation, and iterative correction loops.

Accomplishments that we're proud of

  • Fully automated paper-to-JSON lab template conversion
  • Native multimodal understanding of medical documents
  • Auto-generated medical formulas and conditional logic
  • Live interactive refinement via chat
  • Production deployment on Vertex AI with Gemini 3 Flash

What we learned

  • Gemini 3 Flash excels at fast structured reasoning when guided with schema constraints
  • Multimodal medical document parsing is now practical
  • Healthcare digitization needs reliability + validation, not just text generation

What's next for SmartCarePlus – AI Lab Report Template Digitizer

  • Expand support for complex pathology and microbiology reports
  • Add multilingual report recognition
  • Integrate FHIR/HL7 export for hospital interoperability
  • Offer one-click onboarding for entire lab catalogs

Our vision is to make healthcare report digitization instant, universal, and error-free.

Built With

Share this project:

Updates