Info

  • Category : Computer Vision, AI/Machine Learning
  • Date : February, 2018
  • Client :Axis Bank
  • Create templates for each form type
  • Character recognition
  • Apply transformations to input image
  • Prevent not-first-time-right
  • Identify separate form fields
  • Removing noise and other elements
  • Optimise OpenCV for form processing
  • Faster than manual maker-checker
  • Check if field is filled or not filled
  • Scalability
  • Remove false-positives
  • Long-term cost effectiveness

Step 01

Understand requirements & architect it

Our experts analysed the system and processes involved with eKYC form processing within Axis Bank end-to-end to be able to come up with a solution that will be able to replace the manual form processing efforts involved. We then went ahead with a high-level architecture and design breakdown highlighting the tools, techniques and flow of the SuD (System under Development).

Step 02

Begin Development

The next step was to develop the SuD taking into account limited data since Axis Bank is a financial institution. We had access to 4 forms and had to design our solution to be fault-tolerant at scale. There were various intermediary challenges that arose due to limited upfront data. However, we were able to overcome those challenges and helped Axis Bank innovate.

Step 03

Deploy & Scale

A thorough security audit was performed as the SuD would deal with 3 Lakh forms every month at least containing private and confidential customer data. Once everything was passed, we setup the system on the servers provided by Axis Bank while accounting for accuracy and scalability of this solution. This was a challenge since AI/ML techniques take a lot of compute. But we were able to overcome and succeed.

Interested To Get Our Featured Service