On a typical day, mobility aggregator Rapido processes millions of ride requests on its platform—across bikes, autos, and cabs—peaking at 6.5 million.
Behind that scale is a distributed architecture designed to match riders and drivers in real-time while navigating the complexities of Indian cities.
In an interview with AIM, Rapido CTO Srivatsa Katta explains that the platform’s highly distributed backend system enables captains (drivers) to choose from a plethora of rides within seconds.
“It’s a very complex system with hundreds of microservices, fully cloud-based,” he says, adding that machine learning systems and low-latency infrastructure are critical to quick matching rides at this scale.“Every architectural decision we make is about resilience and scalability. There are no single points of failure—everything runs in clusters, has backups and scales automatically based on load.” Katta adds.
What Happens When You Book a Ride
Every ride request triggers a series of backend processes designed to quickly identify available drivers and determine whether a trip is feasible.
When a user books a ride, the platform runs several geographical queries to locate nearby drivers and estimate the best possible match. The system then sends ride requests to drivers along with information such as pickup point, drop location, and expected earnings.
“We give the choice to the captains whether to accept or not,” Katta says. Drivers can view the trip details before deciding on a ride, which he says helps maintain transparency in the marketplace.
Reducing the time it takes drivers to accept a ride and reach passengers remains one of Rapido’s biggest optimisation challenges. “Where a lot of technology kicks in, [the challenge] is how do we reduce the time for acceptance, time for pickup,” Katta remarks.
Big Data
To make ride dispatch more efficient, Rapido relies heavily on data collected over years of operations. The company has recorded more than 3 billion rides since its founding in 2015, giving it a large dataset that helps improve route prediction, ride matching, and estimated arrival times (ETAs).
“That’s a lot of data that we have, which we use with machine learning and data systems to figure out what a real ETA looks like,” he says.
This historical data is layered on top of real-time traffic signals from mapping providers.
“It’s a combination of what some of the map providers give us, real-time traffic information, distance, time—all of it,” Katta adds.
The company also train many engineers to better use AI tools and the ecosystem to improve productivity.
Mapping India’s Unpredictable Roads
Building reliable routing systems in India presents unique challenges compared to other markets, with road closures due to construction and evolving traffic patterns making real-time navigation complex.
Even official mapping rules sometimes differ from real-world driving behaviour. Katta says drivers may take turns or routes that map systems technically classify as restricted.
“By the book, when the road has a solid line you’re not supposed to cross,” he says. “But in India, that line is also not visible… people can actually take a right turn.”
Because of this, Rapido does not rely on a single mapping provider. Instead, the company combines multiple data sources with its own ride data to improve routing accuracy.
“We use a combination of cloud providers as well as open source systems like OpenStreetMaps”, he adds.
Accepting Rides
Although algorithms help identify which drivers receive ride requests, the final decision to accept a trip remains with the driver.
“Most of them look for economic feasibility of the ride,” the CTO remarks.
Drivers also consider factors such as pickup distance and drop location, particularly during peak traffic hours. Some drivers may avoid congested areas where traffic delays reduce the number of trips they can complete. Others may avoid locations where the chances of getting another ride after the drop are low.
“Some people look for the drop location… they don’t want to go into very crowded areas or places where the likelihood of getting the next ride is very low,” Katta says.
Different Model for Autos and Cabs
Beyond the technology platform, Rapido has been experimenting with different marketplace structures for its mobility services.
For auto-rickshaws and cabs, the company has introduced a software-as-a-service (SaaS) model, where drivers pay a small daily fee to access ride demand on the platform. “Basically, for autos and cabs, there is a very small amount they pay for a day to discover the demand,” Katta explains.
Under this model, Rapido acts primarily as a discovery platform connecting riders with drivers rather than controlling pricing directly. “We just make sure that demand is made visible to the captains,” he says.
Drivers and passengers then decide how the ride proceeds, including the final route or any adjustments during the trip.
Bike taxis on the platform, however, still follow a commission-based aggregator model, where Rapido takes a small percentage from each ride.“It’s mostly on a commission-based model, which is a typical aggregator model,” Katta says.
The bike taxi segment, which has been central to Rapido’s growth, has also faced regulatory challenges in several Indian states.
In Karnataka, authorities have moved to restrict or ban bike taxi services in recent years due to licensing and regulatory concerns. Maharashtra also recently revoked provisional licenses for bike taxis.
Such regulatory uncertainty has forced mobility platforms like Rapido to continuously adapt their business models and service offerings.
Expansion across cities
Katta says the company now launches services in 30–40 cities every month, a pace made possible by building highly configurable technology platforms.
Much of the platform infrastructure is designed so that services can be activated simply by configuring specific features for each city rather than building new systems from scratch.
“The technology teams rarely get involved in city launches,” he adds.
The company is increasingly focusing on expanding into tier-3 and tier-4 cities, where organised ride-hailing services remain limited.
“There’s 300+ million mobility happens every day,” Katta says, indicating the market space for ride-hailing services.
In addition to improving transportation access, Rapido also sees its platform as a way to create flexible earning opportunities for drivers in smaller cities.
“Rapido, being an on-demand platform, means you can choose how many hours you want to work,” he says.
Chasing that scale also reflects on Rapido’s topline. The company’s revenue from operations grew 44%—from ₹648 crore in FY24 to ₹934 crore in FY25—with most of it coming from platform services, but subscription income and delivery services also contributing significantly. It also narrowed its losses by over 30% to ₹258 crore.
Your reaction