ALIMAAZ
Back to Projects

GLR Pipeline Automation

AI AutomationLLMsFastAPIOCRNext.js

Guideline Loss Report (GLR) Claims Portal — a full-stack automation web application that extracts information from PDF reports using LLMs and dynamically fills Microsoft Word template placeholders.

About the Project

Guideline Loss Report (GLR) Pipeline Automation is a full-stack claims processing portal. It enables users to upload a Microsoft Word template containing arbitrary text placeholders and multiple PDF photo reports, contractor logs, or inspection records. The application dynamically scans the Word template for placeholders, extracts matched data from the PDFs using advanced LLMs (Gemini, Groq, and OpenRouter), and allows users to review and download the completed Word document with identical formatting.

Core Processing Pipeline

The system runs an end-to-end automated document extraction and template mapping pipeline:

Upload

Word Template & PDFs

Scan

Extract Placeholders

AI Extract

LLM Document Mapping

Verify

Interactive Review

Download

Formatted Word Doc

Key Product Capabilities

Built to deliver high accuracy, low latency, and a premium inspection workflow:

Tri-Provider LLM Selector: Choose between Google Gemini (direct structured JSON inference), Groq (low-latency Llama models), and OpenRouter (GPT/Claude fallback).

Keyless User Mode: Server-level preconfigured API keys mean users can run extractions without entering credentials.

Interactive Verification Editor: Dashboard featuring live field completion badges, dynamically grouped input fields, and a scanning radar overlay.

Clean-up Background Tasks: FastAPI background tasks automatically purge temporary uploads and compiled outputs from the server after download.

Tech Stack & API Layers

Harnessing async APIs and clean typescript components:

Frontend: Next.js (App Router), React, TypeScript, Lucide Icons, Vanilla CSS (dynamic drop-zones, stepper progress indicators, radar scanning effects)

Backend APIs: FastAPI (Python REST API), python-docx (Microsoft Word manipulation), PyMuPDF (PDF text reader), Pytesseract OCR (image text parser), Pillow (image conversion)

AI Orchestration: Gemini-3.5-Flash (structured mapping), Llama-3.1-8b via Groq (speed-optimized inference), GPT-3.5-Turbo via OpenRouter (fallback options)

Project Demo

Project Note

⚡ Accelerating Property Insurance Claims Historically, claims adjusters and property inspector teams spend hours transcribing details from physical PDF reports into Microsoft Word templates. This platform automates the pipeline entirely using specialized LLMs with custom paragraph run-merging to protect document styles.