AI Document Processing Assistant — Portfolio Case Study

Live demo: https://stebenix.github.io/ai-document-processing-assistant/

An AI Integrator portfolio case study for automating document-heavy finance operations — from document intake and OCR/LLM extraction to validation rules, human review, audit trail, and ERP-ready export.

What this project shows

• AI document intake from email, DMS and PDF upload
• OCR + LLM-style field extraction workflow
• Validation rules for IBAN, VAT, PO matching and ERP export readiness
• Human review queue and audit-ready governance logic
• Integrations screen with connected systems and connector health
• Analytics dashboard with operational KPIs
• Responsive desktop, tablet and mobile navigation

Tech stack

• Vite
• Vanilla JavaScript
• HTML/CSS
• Responsive dashboard UI

Run locally

npm install
npm run dev

Then open the local URL shown in the terminal.

Build for production

npm run build
npm run preview

The production build is generated in dist/.

Deploy

This project can be deployed to GitHub Pages, Vercel or Netlify.

GitHub Pages auto-deploy

A ready GitHub Actions workflow is included in .github/workflows/deploy.yml. After pushing to the main branch, enable GitHub Pages in repository settings and select GitHub Actions as the source.

For Vercel/Netlify: import the GitHub repository and use:

• Build command: npm run build
• Publish directory: dist

Python automation engine

This project includes a Python simulation engine that demonstrates the operational logic behind the dashboard:

• document intake
• OCR/LLM-style field extraction simulation
• validation rules
• risk scoring
• human review routing
• audit trail
• ERP export readiness
• AI recommendations

The frontend is a static product demo. The Python engine simulates the business logic and generates structured output data for the case study. Python does not need to run on GitHub Pages; it writes outputs/demo_results.json as static evidence and refreshes public/demo_results.json as the Vite/GitHub Pages-compatible static frontend copy.

Static data bridge

The Python engine writes outputs/demo_results.json. For the static GitHub Pages dashboard, selected summary values can be exposed to the frontend through a static JSON copy, allowing the UI to display pipeline-generated metrics without a backend runtime. In this project, outputs/demo_results.json is the Python-generated source and public/demo_results.json is the static frontend copy loaded by the dashboard.

Intended refresh flow:

python -m python_engine.main
npm run build

Run it locally:

python -m python_engine.main

Run tests:

pytest

The automation layer is intentionally lightweight and portfolio-friendly: standard-library pipeline modules, deterministic sample data, explainable validation/risk logic, and clean JSON output without databases, APIs, Docker, Flask, FastAPI, or heavy ML dependencies.

Portfolio note

Portfolio case study by Oleksandr Stebeniev · 2026.