AI Home Loan Eligibility Agent
Budget: ₹1,500 – ₹12,500 INR
I need a basic prototype that lets me upload a customer’s PDF or JPG paperwork and instantly returns a below
1. Loan Eligibility
2. Pending Documents (showing the document type only)
3. Pending Form Details
4. Probable Queries the customer should expect
I will provide the full rulebook—eligibility criteria, a master document checklist, all form templates, and a log of past queries. The system must read those references once, store them, and use them as ground truth for every new customer file it analyzes.
Key points
• Input formats: PDFs and JPGs only.
• Eligibility logic must put Income details first when scoring.
• The “Pending Documents” section should list the missing document types exactly as they appear in my checklist. No extra metadata is needed.
• Accuracy and clear, structured output are critical; if the agent cannot classify a page, it should flag it for manual review rather than guess.
Tech flexibility is welcome—Python with OCR (Tesseract), a lightweight embedding model, or any comparable approach is fine as long as setup is straightforward on a standard Windows machine.
Deliverables
• Working script or small app that accepts uploads and produces the report.
• Read-me explaining installation, configuration, and how to update the rules or reference lists.
• A quick demo run using sample files I’ll share.
If you’ve built similar document-driven eligibility tools before, let me know; otherwise, outline the libraries you plan to use so I can green-light them quickly.
1. Loan Eligibility
2. Pending Documents (showing the document type only)
3. Pending Form Details
4. Probable Queries the customer should expect
I will provide the full rulebook—eligibility criteria, a master document checklist, all form templates, and a log of past queries. The system must read those references once, store them, and use them as ground truth for every new customer file it analyzes.
Key points
• Input formats: PDFs and JPGs only.
• Eligibility logic must put Income details first when scoring.
• The “Pending Documents” section should list the missing document types exactly as they appear in my checklist. No extra metadata is needed.
• Accuracy and clear, structured output are critical; if the agent cannot classify a page, it should flag it for manual review rather than guess.
Tech flexibility is welcome—Python with OCR (Tesseract), a lightweight embedding model, or any comparable approach is fine as long as setup is straightforward on a standard Windows machine.
Deliverables
• Working script or small app that accepts uploads and produces the report.
• Read-me explaining installation, configuration, and how to update the rules or reference lists.
• A quick demo run using sample files I’ll share.
If you’ve built similar document-driven eligibility tools before, let me know; otherwise, outline the libraries you plan to use so I can green-light them quickly.