India Document Tools — Form Processing & OCR Platform
Full-stack document intelligence platform combining Indian exam/application form preparation utilities with OCR-backed structured extraction
The Problem & Target Context
Indian applicants face stringent, fragmented upload constraints across exam and recruitment portals (exact pixel dimensions, file size clamps, signature ratios), while manual verification of marksheets and identity cards creates massive administrative friction.
Engineering Approach & Core Principles
Built a multi-tier microservice architecture separating client-side image and PDF manipulation utilities from queue-backed FastAPI/PaddleOCR extraction workers with confidence-scored structured outputs.
What I Built & Individual Contributions
Designed the Next.js 14 user interface, Fastify backend API gateway, Redis job orchestration, PostgreSQL persistence via Prisma, and integration contracts for the Python PaddleOCR extraction service.
Architecture & Data Pipeline
User Document Upload → Fastify API Gateway → Redis Job Queue → Python PaddleOCR Microservice → Structured Key-Value Extraction → Next.js Verification Dashboard
[ Applicant Document / Form Upload ]
│
▼
┌─────────────────────────────────────────────────────────┐
│ Fastify Gateway & API Service │
│ ├── MIME/Dimension Constraints & Presets │
│ ├── Redis Distributed Job Queue │
│ └── PostgreSQL / Prisma Metadata Store │
└────────────────────────────┬────────────────────────────┘
│
▼
┌─────────────────────────────────────────────────────────┐
│ Python PaddleOCR Extraction Service │
│ ├── Document Classifier (Identity, Marksheet, Invoice)│
│ ├── Bounding-Box Key-Value Extraction │
│ └── Confidence-Scored JSON / Markdown Emitter │
└─────────────────────────────────────────────────────────┘Key Engineering Decisions & Trade-Offs
Separated high-throughput client-side file manipulation (canvas image resizing, client PDF optimization) from server-side OCR workloads. Isolated the Python PaddleOCR extraction engine behind a Redis job queue with Fastify gateway rate limiting and Prisma persistence.
Testing, Verification & Reproducibility
Multi-tiered test framework: Pytest for Python OCR parsing, Node.js integration tests for Fastify endpoints, and Playwright end-to-end browser automation workflows ('npm run test:e2e').
Verified Outcomes & Metrics
Sub-second client-side image preprocessing; automated table and field extraction; automated Playwright E2E test coverage; zero cloud telemetry leakage.
Verified Implementation
Verified multi-tier test architecture: Pytest for OCR extraction logic, Node.js integration tests in backend, and Playwright end-to-end browser automation workflows. Verification here represents code audit and local inspection, distinguishing functional code paths from external cloud provider accounts.
Boundaries & Known Limitations
Independent software utility; not affiliated with or endorsed by any Indian government entity or examination board. Extraction accuracy depends on document scan resolution.
Related Portfolio Projects
Explore companion systems in similar domains across AI engineering, quantitative finance, and developer infrastructure:
Collaborating on AI Engineering or Quantitative Systems?
I design and build end-to-end production software, quantitative models, and reliable AI architectures. Let’s discuss technical challenges or product collaboration.