AI PDF Text Validator Tool

Job ID: 39572621

Budget: £750 – £1,500 GBP

Project Overview:
We require the development of a desktop application that can intelligently compare up to four mortgage-related PDF documents side by side. These PDFs may contain both text-based and scanned/image-based content. The software must extract key information (via OCR where needed), interpret and normalise variations in field names, and utilise a local AI model running in LM Studio to assess and highlight matches or discrepancies.

The output must be a clean, professional PDF report summarising agreed key fields and custom sections like “Special Conditions.” All processing must be fully offline and GDPR compliant — no external API usage is permitted.

Key Requirements

1. Offline Operation / GDPR Compliance
Must operate 100% offline

All LLM (AI) processing must route through LM Studio running locally on the same machine via http://localhost:1234

No internet connection or cloud API usage allowed

2. User Interface (GUI)
Simple interface allowing the user to:

Select up to 4 PDF files

Click a button to "Compare Documents"

View results in a scrollable UI and optionally export them to a PDF report

GUI must handle blocking operations in background threads to avoid freezing

3. Document Handling
Must accept any combination of:

Text-based PDFs

Scanned/image PDFs

Password-protected PDFs (optional bonus)

Must automatically perform OCR where needed (using Tesseract or similar) to make all content readable for AI

4. Key Field Comparison (via LLM)
Use LM Studio's OpenAI-compatible API to process the documents locally using a hosted model (e.g. LLaMA3 or Mistral)

Pass extracted text to the LLM along with a structured prompt to compare specific fields such as:

Name
Address
Loan amount
Term
Fixed period end date
Expiry date
Purchase price
Valuation
+others (editable list)

5. Field Variations / Help Dictionary
Allow a local dictionary or alias list to assist the AI with variations in field labels (e.g. “Loan Amount” = “Amount Borrowed” = “Principal”)

Ideally, make this list user-editable (e.g. via JSON or embedded GUI editor)

6. Pre-defined Section Extraction
Extract and include free-text sections from the PDFs where found, such as:

"Special Conditions"

"Additional Notes"

This should be editable fields also. These should be clearly labelled and included in the final report and LLM output

7. PDF Report Output
Final output must be saved as a clean, printable PDF

The report must include:

Side-by-side or tabular comparison of all specified fields

Clear indication of matched and mismatched values

Sectioned output of any “Special Conditions” or notes from each document

Document filenames and timestamps

Deliverables
Full working Python project (including GUI, LLM interaction, OCR, and report output)

A user guide / README

Sample output report (PDF) using dummy data

Optional: Installer or packaged .exe (bonus)

Ideal Freelancer Will Have:
Strong Python development skills

Experience with OCR and PDF parsing

Familiarity with local LLMs (especially LM Studio)

Awareness of GDPR constraints in software design

Strong communication and ability to deliver well-commented code

Additional Notes
If LM Studio API changes or is unavailable, the system must fail gracefully and notify the user