K

Kestrel Consumer Finance

Automated loan review — demo

K

Kestrel Consumer Finance

Automated loan review — demo

checking platform… signed in as

1. Upload a loan application

Drop a PDF loan application below. It's read by the loan-company agent's read_document step, which asks a model to extract the applicant's details as structured JSON — nothing here decides anything, it only proposes what the form says.

📄

Click to choose a PDF or drag one here

Only .pdf, up to 10MB

Or try a sample application:

2. Pipeline

  1. Document received
  2. Extracting fields via model
  3. Applying lending policy
  4. Explaining the decision

3. Result

—

Extracted application

Computed ratios

Policy thresholds

Max debt-to-income35%
Max loan-to-income50%
Document text read from the PDF 

      


      

        Run trace (run )
        

      


      
    

  


  

    

      Extraction is a model — fallible by nature. The decide step is
      plain deterministic Python; the model never sets the outcome, only proposes what
      the document says and, afterwards, phrases it in a sentence.