BharatPe, an Indian fintech firm that caters to local grocery shops, small merchants and consumers, is using the Google Cloud Platform (GCP) for real-time data analytics.
It is using Google’s BigQuery, among other products, Sanjeev Singh, BharatPe’s head of data platform and engineering, said at a fireside chat organised by Google Cloud.
BigQuery is a fully-managed, serverless data warehouse that enables scalable analysis over petabytes of data.
BharatPe offers products such as interoperable QR codes and support for payments via India’s United Payments Infrastructure (UPI).
Singh said his organisation has “millions of users daily” and apart from providing QR codes for UPI transactions, it also arranges loans for small merchants and retailers.
He added that the company has also moved into the peer-to-peer (P2p) space with the launch of its 12% Club app that allows consumers to invest and earn up to 12 percent annual interest or borrow at a similar rate.
Singh said real-time data analysis is an important requirement to build “our data pipeline and data warehouse and the success of an analytic tool depend on the real-time analysis”.
Explaining why Singh observed that real-time data is crucial to understanding “our customers and the transactions going in our systems” as a payment firm.
“It is necessary to have an eye on credit scoring, anomaly detections where we can understand the transactions between different banks and also ensure that we are checking every transaction every minute, every hour daily”, Singh said.
He added that BharatPe has a dashboard running on the real-time system to check active merchants, and reject the sudden spin up or spin down transactions with help of GCP solutions.
Data access
Singh said GCP’s notable feature that allowed data access across various team members performing different tasks was “very useful”.
“We have very specific success criteria because not only the analytics team or data science team access the data; there are several people across the organisation doing analysis and reporting,” Singh said.
He added that the company uses BigQuery to transfer data from one cloud to another in a seamless manner.
Singh said for security reasons BharatPe wants its data to be located within India and that has been possible with the use of GCP.
Additionally, he added, the GCP solutions did not need any upskilling as most of the users in the company were “already quite familiar with the SQL queries” which are quite similar to BigQuery.
The BigQuery machine learning (ML) time series modelling pipeline includes multiple modules and “we are still exploring the time series and ML steps”, Singh said.
Giving some examples, he said BharatPe uses GPC’s Document AI, Vertex AI and Dataproc services to boost its artificial intelligence (AI) and ML capabilities.
“We are using Document AI for document scanning and extraction of data, Vertex AI for image processing and Dataproc for processing of huge amounts of data,” Singh said.
