Due to stringent regulatory requirements and due diligence in the Financial world, customer on-boarding process has become increasingly complex and important too. Unless the process is flawless from start to finish, it can make or break a user’s impression of the the client company. The front-office and operations team of the client are facing 2 major challenges:

1. Customer drop-outs when customers are asked to: fill lengthy forms, in-person verification and produce physical ID documentations. “How do we improve user registration rate to 85%”?
2. The current onboarding process takes too long as it is tedious and involves 15+ manual steps, lots of data entry and physical verification of IDs. “How do we do eKYC in days/minutes instead of weeks”?

Oneture as the prime solution provider worked with client’s Infra & Business team and AWS Professional Services to implement AI models for PAN Card, Aadhar Card and Passport to extract the needed fields after scanning the image as well as on performance and cost optimisations to reduce cost-per-KYC and customer on-boarding

Oneture team with expertise in AWS Services like EC2, S3, SageMaker, deep learning models e.g. YOLO architecture and framework implemented end to end solution which includes web application in Angular Framework and integrated AI models.

The proposed approach and solution is based on advanced Digital Technologies like AI/ML, Cloud.

The customer initiates self-onboarding by submitting the Proof of Identity (POI) and Proof of Address (POA) e.g. PAN, Aadhar Card, Passport via an app or a web portal.

These documents are then sent to the Amazon SageMaker endpoint to analyse and extract relevant information from the images, e.g invoke the “passport-validation” endpoint using the AWS Python SDK.

The underlying ML model validates user’s passport image and extract relevant user information from the image for data capture verification process. E.g.  country, date of birth, expiration date, nationality, sex, name, and surname from the passport image.

In few cases cross-validation is done manually, agents re-validate the data or take a second look at validations rejected by the algorithm.

Based on this extracted information and minimal data-form filling, the Solution displays the same on the screen for customer’s review and final submission.

In future, we are planning to implement AI/ML driven video-ID-based KYC to take care of authenticating the document and identity of the customer by matching the recorded video and documents submitted.

The benefits of digitising the ID collection process are several, the most important of which is that onboarding time is decreased, fraud risk is lowered, and even the most stringent ID and verification requirements are met promptly.

• Oneture team demonstrated a quick learning curve to get hands on for picking up the work AWS Proserve team had started and implementing the AI models within the timelines with 85%-95% accuracy
• The team trained 4 AI models on thousands of PII documents in a span of two months which once integrated into the pipeline can reduce the steps needed and time taken for NRI registration for the Client 
• Product development in agile methodology to deliver incremental updates to the application, with constant stakeholder feedback and updates to the user journey to achieve client's vision.
• The solution takes care of the In-Person Verification (IPV) mandated by SEBI and RBI, meanwhile saving on operational costs and time that comes with paper-based KYC

• Being agile and having a good learning curve is important
• Maintaining a collaborative atmosphere to ensure consistent delivery is crucial
• Keeping in mind the interests of both customer and technology partner helps in smooth delivery of the project 
• Maintaining the right balance between feedback, course corrections, and deliverables.