Home/Case Study: Computer Vision and Text Mining in FinTech
Case Study: Computer Vision and Text Mining in FinTech
Fintech Enjoys a Fifteen-Fold Decrease in Time Processing Bank Statements and Increase in Extracted Text with AI Use
Kashfia Mahmud, CEO of a fintech startup, was hitting an operational bottleneck as her product’s popularity grew. The new peer-lending platform, based in a developing economy, needed to vet borrowers’ creditworthiness for its lending partners without access to electronic bank statements. The company was manually analyzing paper statements but as the operation scaled, this process was quickly becoming a source of friction. The CEO wondered if Infolytx could build an automated solution that would intelligently analyze the bank statements prospective borrowers submitted online.
The Infolytx team had recently launched its Deep Vision™ and Text Miner™ Accelerators which combined could solve the company’s problem. Led by the Infolytx CTO, the team rapidly architected a solution and API service. Machine and Deep Learning models were developed to determine whether submitted paper document images were authentic bank statements. Next, the Infolytx Text Miner™ Accelerator and Natural Language Processing capabilities were applied to extract and infer relevant information from the text even when the incoming sources were poorly scanned or photographed documents. The information extraction included applicant’s names, bank names, addresses and postal codes, as well as tabular extraction of transaction data. Additional information was inferred, such as the existence and amount of installment loans, salary data, and other transaction-related information.
In collaboration with the client’s financial analysts, Infolytx engineers trained appropriate Machine and Deep Learning algorithms to decrease the statement processing time from 15 minutes to under one minute – a fifteen fold increase in efficiency. Furthermore, the additional extracted information available for review was a significant factor in enabling better credit decisions. “This is the type of impact Infolytx aims to deliver,” noted Badrul Husain, Infolytx CEO.