KYC automation using Azure AI

Overview of My Submission

Leveraging Azure AI form recognizer cognitive service to automate customer KYC verification and validation process. Using this project submission, manual reviewers do not need to analyse and go through each and every submitted document by the customers. The AI service will reject invalid and bogus KYC documents.

👉 Lets have a brief knowledge about KYC and legacy KYC verification flow.

What is KYC?

Businesses need to ensure customer due diligence (CDD) using know your customer (KYC) documents. CDD involves verifying that customers are who they say they are and assessing the risks like the possibilities of fraud, money laundering, terrorism financing etc.
KYC process includes verifying customer’s name, address, photograph by analysing bank documents, utility bills, etc.

Legacy KYC process 🛠

1. Customer need to upload his KYC document like PAN, Utility bills etc. through an online portal.
2. Manual reviewer reviews each and every document uploaded by customer and verifies the following checks:

• Integrity of uploaded KYC document.
• The image quality of KYC document.
• Validating information(Name, Father Name, DOB, Document number) in KYC document.

Azure AI based implementation:

1. Customer need to upload his KYC document like PAN, Utility bills etc. through an online portal.
2. After receiving KYC document, custom trained form recognizer model with run and validate:

• Structure (template) of KYC document.
• Extract data from KYC document.
• Then according to business logic, extracted data/ fields from KYC document are verified.

Submission Category:

AI Aces, Computing Captains

Resources Used:

1. Azure AI Form Recognizer (For custom machine learning model).
2. Azure Functions (For front end api and business logic layer).

Link to Code on GitHub

# API to submit KYC document for analysis. 
curl --location --request POST 'https://pan-kyc-automate-1234009.azurewebsites.net/api/kyc-doc-upload' \
--form 'file=@"jS7VOIp6r/image.png"'

# After submitting document in step 1, you will get requestId in response, You have to use that with other informations that you want to validate.
curl --location --request POST 'https://pan-kyc-automate-1234009.azurewebsites.net/api/kyc-doc-verify' \
--header 'Content-Type: application/json' \
--data-raw '{
    "name": "D MANIKANDAN",
    "fatherName": "DURAISAMY",
    "pan": "BNZPM2501F",
    "dob": "16/07/1986",
    "requestId": "3c89950b-d282-46c1-9c54-fb565fc1ab20"
}'

Screen shot

Customer submitting KYC documents

Screen after waiting for KYC repsonse

Architecture:

References:

• https://eastus.dev.cognitive.microsoft.com/docs/services/form-recognizer-api-v3-0-preview-1
• https://docs.microsoft.com/en-us/azure/azure-functions/
• https://azure.microsoft.com/en-in/services/storage/blobs/#overview
• https://azure.microsoft.com/en-in/services/form-recognizer/