Founded in 2014, food ordering and delivery platform Swiggy has expanded its operations to over 700 cities in India. More recently, it launched a live events ticketing service called Swiggy Scenes. To support this rapid business growth, the company partnered with Confluent to simplify complex infrastructure operations.
Akash Agarwal, Architect at Swiggy, spoke with Frontier Enterprise during the Confluent Current 2025 conference in Bangalore about their tech stack, expansion plans, and how Confluent is helping address their internal challenges.
Could you share a bit more about what Swiggy does?
We started as a food delivery app, and now we also handle groceries and other business lines. What we’re trying to solve is the need for a true hyperlocal commerce delivery experience.
It’s not just about connecting the consumer, restaurant, or the delivery partner. We want to give each of them a reliable, transparent, and unique experience. That’s what sets Swiggy apart.
To deliver on that, we need sub-second latencies — and that’s where we leverage Kafka.
What were you using before Kafka?
I joined in 2019, and at that time, we were using open-source Kafka. Kafka is a great tool, but at that time, we were scaling heavily, and once we reached that scale, using Kafka wasn’t very straightforward. We need a team to manage these systems and tweak them to handle scale-up and scale-down scenarios. There was a lot of maintenance overhead.
That’s when I felt the need to start exploring what other options were available. We eventually decided to go with Confluent Cloud because they were the best fit. They’ve helped us manage the complexities on the infrastructure side, which means we can now concentrate much more on the business side.
We can implement governance and invest our energy in the business, because at Swiggy, we believe that just because we can do something doesn’t mean we should.
We don’t want to be an infrastructure company. We don’t want to reinvent the wheel or build expertise in something that isn’t our core focus. Instead, we’d rather build expertise in our core business — developing algorithms or new things that make our systems faster, simpler, and more cost-effective.
Could you give an overview of your infrastructure stack?
We use AWS, Confluent, Databricks, Snowflake, etc.
Last year, the CTO of Confluent was saying that Confluent Cloud is actually 16x faster than Apache. Have you experienced that giant increase in speed?
I can’t really comment on the speed. Over the course of our journey, we’ve definitely seen significant improvements, but whether it’s faster than open source or not — I think Confluent would have done their own benchmarking for that.
For us, Confluent takes care of all the version upgrades, security-related stuff, all the compliance, so now we just have to take care of our business.
Confluent also offers user-based clusters. In our case, we have a few high-peak scenarios — like New Year’s Eve — where we get much more traffic than usual. We need to scale up and down during events like that, and Confluent Cloud really helps. They’ve made upscaling and downscaling much simpler.
We used to invest a lot in this, and it was only semi-automated. Now, it’s just a one-click operation.
What’s the most exciting thing happening over at Swiggy’s labs?
Right now, we’re focusing on the AI side of Swiggy — and not just AI for the sake of it, but building an open environment where both business and technical teams can work more effectively using these tools. We’re also looking at how we can use AI to improve the experience for all our main actors, like customers, restaurant partners, and delivery partners.
What particular AI use cases are you looking at right now?
We’re working on multiple innovation initiatives with large language models, like chatbots and similar applications. It’s not just about a few use case integrations — we’re also thinking about how to use them within the organisation. How can we simplify things for internal stakeholders? How can we work more efficiently using these tools?
You mentioned the AI deployment on the business side. Could you talk about the use case specifically?
Let’s say there’s a report a business user needs to look at before making a decision. Instead of going through the entire report, that person can now chat with it — interact with the data directly. That not only helps retrieve the necessary data, but also provides suggestions from the AI.
