Spotting faulty signatures on bounced cheques using computer vision and ML

The algorithm developed by tech entrepreneur Mahadevan Jayaram's firm has been tested on more than 20,000 handwritten signatures to establish their genuineness

A consummate marketer-turned-technology entrepreneur, Mahadevan Jayaram is trying to find out solutions to an age-old problem faced by banks -- cheque clearance.

Any bank officials involved in the clearance process will tell you about the precision involved and the repercussions of clearing a faulty cheque. Jayaram, co-founder of Chennai-based DeepQuanty Artificial Intelligence Labs, is trying to automate the process which will not only save a lot of costs for banks, but also will help in faster clearance.

"We use computer vision to extract data from a cheque, which then is processed through machine (ML) learning, before giving the go-ahead for clearance," says Jayaram. "Currently, we use ML to extract data in the BFSI space as it is seen as the low-hanging fruit. Going ahead, we will tap other industry verticals such as retail, ecommerce and healthcare among others."

DeepQuanty's first banking product, SnapChek, can read handwritten and printed text on a cheque."Reading printed cheques is easy with OCR (optical character recognition) technology. However, reading handwritten ones is tricky, but our product is able to do so with 85 per cent accuracy," Jayaram says.

According to the company, the machine learning-based algorithm has been tested on more than 20,000 handwritten signatures to establish genuineness. As the machine is trained with more data points, the accuracy will also increase.

When asked about the falling usage of cheques, which may limit the growth of the firm, Jayaram says around 90 per cent of business transactions continue to be conducted using cheques, which are mostly for treasury operations. Reserve Bank of India data showed that cheques account for only three per cent of the banking system's total transaction, and this is likely to come down further in coming years. "Use of cheques is falling at a slower pace in the financial system. So, there is plenty of opportunity left in this space," Jayaram says.

DeepQuanty, which started operations in 2018, is expanding its offerings to other areas in order to bring scale, apart from having a cushion against a slowdown in any segment. For instance, it is planning to come up with solutions to read various application forms used in the financial system. "In the case of current account or savings account opening, the whole application is handwritten by the applicant. Currently, we are working on a product that will be able to read the complete application accurately," Jayaram says. This product will require training with a lot of handwriting samples to extract data points from an application and tabulate it in a desired manner for verification, he adds.  

While DeepQuanty is working on various innovative AI-based solution, Jayaram is not alone in this journey. While his previous work experience as brand manager at HDFC Bank and VP-product development at Bharatmatrimony are coming in handy, Jayaram has the support of Jayaram K Iyer, another co-founder who was earlier the chief strategy & analytics officer of Matrimony.com. "JK (Iyer) and I were independently doing something in the computer vision space during the 2015-2017 period. Then, we decided to come together in 2018 based on our common vision," Jayaram said.

Currently, DeepQuanty has a 20-member team in Chennai of which 16 are engineers working in computer vision and data analytics. The company is now looking at opening a centre in Bengaluru.