Linux
Docker
AWS
Python
JavaScript
TypeScript
C++
Java
HTML5
CSS3
SQL
React
Next.js
FastAPI
Node.js
Flutter
PyTorch
Scikit-learn
Pandas
NumPy
MongoDB
Git
Figma
fromVedangimportCreativity
Back to Selected Work
MEDICAL AI2026

Swasthya Saathi

Medical Report Intelligence Platform

Medical report intelligence platform that converts PDF and image medical reports into accessible health insights using OCR and NLP.

Swasthya Saathi — project screenshot
Category
MEDICAL AI
Year
2026
Core Stack
React · FastAPI · Python · MongoDB · Tesseract OCR · OpenRouter · Ollama

Project Context & Purpose

Swasthya Saathi bridges the gap between clinical jargon and patient comprehension. It processes PDF and image reports, extracts biomarkers with OCR and rule-based parsing, and presents results in plain language with contextual explanations.

Swasthya Saathi uses a React frontend and FastAPI backend. The app supports PDF and image uploads, then runs OCR, parsing, and analysis to extract report values.

Technologies & Infrastructure

ReactFastAPIPythonMongoDBTesseract OCROpenRouterOllama

System Design & Workflow

Tesseract OCR handles digitization, then regex and spaCy-based parsing extract parameter names, values, and reference ranges. A rule-based analysis layer returns structured JSON, and the backend can also provide optional LLM explanations.

Engineering Challenges & Solutions

CHALLENGE 01Resolved

Handling report formats from different labs and hospitals with inconsistent layouts

CHALLENGE 02Resolved

Distinguishing numerical values from reference ranges in unstructured OCR output

CHALLENGE 03Resolved

Building a medical NLP pipeline without proprietary datasets

CHALLENGE 04Resolved

Designing an accessible, anxiety-reducing UX for patients receiving critical health data

Making medical reports easier to understand

The app is designed to improve health literacy through OCR, deterministic analysis, anonymous sessions, and optional safe LLM explanations.

Upload Types
PDF / Image
Parsing
Rule-based
Explanations
Optional LLM
Sessions
Anonymous