PhonePe has recently developed an on-device machine learning (ML) framework, called the Edge Framework, designed to process user data directly on users’ phones or tablets.
By handling tasks on-device, PhonePe aims to reduce the dependency on constant internet connectivity, ensuring quicker response times and enhancing data privacy, it said in a blog post.
The fintech giant said the localised processing of sensitive data limits the exposure of personal information to external servers, reducing potential risks associated with data breaches or unauthorised access.
The system can autonomously extract bill details from SMS messages, generating timely payment alerts without compromising financial information. The car insurance self-inspection feature leverages ML to validate vehicle inspection images in real-time, detecting incorrect image angles, identifying incomplete photographs, and providing immediate user feedback that significantly reduces policy rejection rates.
Using advanced optical character recognition, the system can scan credit card details from video frames, recognising formats from multiple providers, extracting card numbers and expiration dates, and eliminating the need for manual data entry, the company said.
It also enables performance optimisation through low-latency processing, eliminating the delays typically associated with cloud-based systems. Operational flexibility allows for rapid feature deployment and modification without requiring complete application updates. Additionally, a continuous improvement mechanism enables machine learning models to become more accurate and precise with each interaction.
Looking forward, PhonePe plans to expand the framework’s capabilities into more sophisticated applications. Future developments include enhancing KYC processes through advanced features like user photo liveliness detection, identification document recognition, and automated correction prompts for improper document submissions.
The Edge Framework maintains a strict consent framework, where permissions are explicitly use-case-specific. This approach ensures users retain complete control over their personal data while simultaneously experiencing enhanced, intelligent mobile services.
Recently in its annual report, PhonePe said it has reduced its customer support workforce by 60% over five years, cutting from 1,100 to just over 400 agents, while seeing a 40X increase in transactions. AI-driven chatbots, utilising generative AI and NLP, now resolve over 90% of customer issues.
Fintech major PhonePe has significantly expanded its hardware footprint by managing close to 7 lakh cores across three data centres in India.
