Document Verifying Software

I am looking for an experienced Full Stack Developer to build a SaaS platform that helps hospitals across Africa reduce insurance claim rejections. The Problem: Hospital insurance claims are rejected every month by Insurance Authorities across Africa due to incomplete forms, missing fields, incorrect tariff codes, illegible handwriting, and missing signatures. The Solution: A secure web application where hospital staff upload a PDF (Pre-authorization, Discharge, or Claims form). The software automatically reads the document using AI-powered OCR, checks it against all Insurance Authority rules, and tells the user exactly what is missing or wrong before they submit it to the insurer. Core MVP Features (Must-Haves) We are building a Minimum Viable Product (MVP) first. Please only apply if you are comfortable building these features: 1. User Management & Multi-Tenancy Secure JWT-based authentication (Login/Signup). Role-based access control (Hospital Admin, Biller, Clinician). Multi-tenant architecture: Each hospital must see only their own data (strict data isolation). 2. Document Upload & Storage Drag-and-drop PDF uploader (React dropzone). Secure file storage (AWS S3 or Azure Blob Storage). File validation (PDF format, max size limits). 3. The "AI Brain" – Backend API Integrations (Most Critical) You will build the backend logic (in Python FastAPI or Node.js) to orchestrate these APIs: API Purpose Key Details Mistral OCR 3 Extract printed AND handwritten text from PDFs $2 per 1,000 pages ($1 via Batch API). Designed for healthcare documents—handwritten notes, nested tables, checkboxes, and multi-page forms. OpenAI API (GPT-4o) Analyze extracted text against Insurance Authority rules Analyze extracted text against Insurance Authority rules Email/SMS Service Send OTPs and notifications SendGrid, Twilio, or similar 4. Asynchronous Processing Pipeline Since PDF processing can take time, you must implement a queue system (e.g., Redis + Bull/Celery) so the user isn't waiting on the screen. Status tracking: Pending → Processing → Completed → Error. 5. Validation Results Dashboard A clean React dashboard displaying detailed results: ✅ Green "Pass" for correct fields. ❌ Red "Fail" with specific text explaining why it failed (e.g., "Missing Patient Signature" or "Service Code out of approved tariff range"). Suggested corrections for each error. Overall risk score (Low/Medium/High). 6. Audit Log & Document History: Complete history of all uploaded documents per hospital. Search and filter capabilities. Tech Stack Requirements (Strict) You must be highly proficient in these technologies: Component Technology Why Frontend React.js (or Next.js) For a responsive, fast single-page application. Backend Python (FastAPI) or Node.js (Express/NestJS) FastAPI is preferred for its native async support and automatic OpenAPI docs; Node.js is also acceptable. Database PostgreSQL Robust relational database for users, hospitals, documents, and results. ORM SQLAlchemy (Python) or Prisma (Node.js) For safe and scalable database interactions. Queue/Workers Redis + Celery (Python) or BullMQ (Node.js) To handle long-running OCR/AI tasks in the background without blocking the UI. File Storage AWS S3 or Azure Blob Storage Secure, scalable PDF storage. Hosting/Deployment AWS / Azure / Google Cloud Containerized deployment (Docker preferred). Insurance Validation Rules (To Be Coded into the AI) The backend must programmatically check these specific parameters: Rule Category What to Check Why It Matters Completeness Are all mandatory fields filled? (Patient ID, Diagnosis Code, Facility Code, Dates) Missing fields = automatic rejection Handwriting Analysis Does the AI detect the form was filled by a single person? Most insurers require 2–3 different parties to fill different sections. Single handwriting = automatic rejection Tariff Verification Does the service code match the country's official published tariff guidelines? Upcoding or incorrect codes = rejection Time Limits Is the claim submitted within the insurer's allowed window (e.g., 7 or 14 days after discharge)? Late submission = rejection Signature Detection Are the required signatures visually present? Missing signatures = rejection Ideal Freelancer Profile ✅ Proven experience building SaaS platforms from scratch (please share live links or GitHub repos). ✅ Deep experience with REST API development and integrating 3rd-party APIs (specifically AI/OCR services). ✅ Strong PostgreSQL database design skills (schema design, indexing, and migrations). ✅ Experience with asynchronous task queues (Redis/Celery/BullMQ) to handle background jobs. ✅ Proficient with Git, CI/CD, and cloud deployments (AWS/Azure). ✅ Excellent English communication skills. ✅ Ability to explain technical concepts to a non-technical client. Bonus (Nice-to-Have): Experience in the HealthTech space. Experience with payment gateways (Paystack, Flutterwave, Stripe). Familiarity with Docker and containerized deployments. How to Apply – Screening Questions To filter out generic proposals, please answer these specific technical questions in your cover letter: How would you architect the document processing workflow? (Walk me through the flow: User uploads PDF → Backend receives file → Mistral extracts text → OpenAI analyzes → Results saved and displayed). How do you handle rate limiting and large file sizes when sending PDFs to external APIs like Mistral OCR? How do you ensure data isolation in a multi-tenant PostgreSQL database? (e.g., Hospital A cannot see Hospital B's claims). Why do you prefer FastAPI over Node.js (or vice versa) for this specific project? Have you integrated OpenAI or Mistral APIs before? If yes, please describe the specific challenge you faced and how you solved it. Deliverables & Timeline Deliverable Description Fully functional MVP Deployed on a live cloud URL (AWS/Azure) with a custom subdomain. Source Code Full repository access (GitHub/GitLab) with clear README for setup. Database Schema ERD diagram and migration files. API Documentation Auto-generated Swagger/OpenAPI docs (FastAPI does this natively). Video Walkthrough 1-hour video explaining the system architecture and how to manage the backend. Deployment Scripts Dockerfile and docker-compose (or equivalent) for easy deployment. Timeline: Expected completion within 8 to 10 weeks from project start. What This Project Is NOT ❌ Not a no-code or low-code project (we are writing real code). ❌ Not a mobile app (web-based MVP first). ❌ Not a full-featured enterprise system (we are building an MVP). Important Notes for Applicants This is an MVP. I want a working, secure, and scalable product. I care more about reliable functionality and clean, maintainable code than fancy UI animations. The hardest part is the AI workflow. If you can make the Python/Node.js backend reliably orchestrate Mistral OCR + OpenAI and handle the queueing system, the rest is straightforward. You must be able to explain architectural decisions in plain English. Clear, frequent communication is non-negotiable. Budget & Payment Terms Budget Range: You can quote depending on your best of knowledge, experience and execution skill. Payment: Milestone-based (e.g., 20% upfront, 30% on API integration complete and on dashboard complete, rest 50% on final deployment). Payment Method: Upwork Escrow (100% protected). About the Client I'm just a solopreneur building a solution to solve a critical revenue problem for hospitals. I have a clear product vision and am looking for an engineering partner who can bring it to life. I value transparency, proactive problem-solving, and honest feedback on what is possible within the timeline. I look forward to reviewing your proposals and finding a long-term technical partner for this exciting project.