AIP Podcast EP 76 – Deploying Agentic AI Applications Across Any Cloud with Defang

This episode’s guest, Prakash Sundaresan, Co-Founder and CEO of Defang, joins host Anne to unpack the growing complexity of cloud deployment in the era of agentic AI. Drawing on nearly three decades of experience across Microsoft, AWS, and multiple startups, Prakash shares what inspired him to build Defang and tackle one of the most overlooked challenges in modern AI adoption: deploying applications securely and efficiently inside customer cloud environments. He explains why agentic applications fundamentally change traditional SaaS deployment models, how data sovereignty and compliance are reshaping cloud strategies worldwide, and how Defang enables developers to deploy agentic apps across clouds using a simple, standardized approach. Tune in to learn how Defang is helping startups and enterprises deploy faster, scale globally, and focus on building great AI applications—without getting lost in cloud complexity.

Follow AIP Affiliate, Defang
Website: https://defang.io/
LinkedIn: https://www.linkedin.com/company/defanglabs/

Follow AI Partnerships
Website: https://www.aipartnershipscorp.com/
LinkedIn: https://www.linkedin.com/company/aipartnershipscorp/
X: https://twitter.com/AIPartnerships

The AIP Podcast is hosted by Anne Cheng, on behalf of the AI Partnerships, a Railtown company

Transcript

SPEAKER_00 0:00

The world of deploying cloud infrastructure is one that is fraught with complexity, despite being one that is most inherent in our digital world, from the largest of enterprises to the smallest of startups. Defining one's cloud strategy is only the tip of the iceberg. Enter DeFine Software Labs and its founder, Pratesh Sundarasan, who is our guest today. In today's episode, we will dive into the complex world of cloud deployment simplified. Hello everyone, and welcome back to our AIP podcast, where we interview the world's emerging CEOs in the world of AI. Now, if you've ever worked on an LLM deployment or an agent deployment, you will know how an agent or model capability can only work within the same cloud infrastructure that it was deployed in and how difficult it is to deploy solutions within customer environments as opposed to simply SaaS-driven models. Here with us today is Prakash Swinderason. Who is looking to address these complexities with a startup, DFANG? Welcome, Prakash, to our show. So glad to have you on board with us. Let's dive right in.

SPEAKER_01 1:22

Thank you, Anne. Great to be here. Thanks for having me.

SPEAKER_00 1:25

Thank you for being here. Pratesh, let's talk about you for a moment. You've worked with some of the world's largest hyperscalers. What brought about you starting DeFi?

SPEAKER_01 1:38

Yeah, uh, it's coming up on 30 years for me in the industry. Um spent a bunch of time at Microsoft building enterprise software in the late 90s and 2000s, uh, also the early days of cloud and Azure. Um and then more recently, I was at AWS leading a couple of their analytics services with you know thousands of customers all across the world. And in between those two, I did a startup, uh, which was eventually acquired by Workday, but it was obviously built on cloud, built on AWS. So I got to see cloud from both sides, you know, from the perspective of those building and bringing those to market, and also from the perspective of a customer, first a small customer as a startup, and then post-acquisition, also a large customer. So that's kind of what led to DFING, some of the challenges we saw in those in those previous uh journeys.

SPEAKER_00 2:35

That's amazing. You know, cloud deployment in an AI native era is increasingly complex, despite having been around for a very long time. Can you tell us why?

SPEAKER_01 2:46

Yeah, uh absolutely. Yeah. You know, cloud deployment, as you alluded to, was a challenge, is a challenge even before the advent of AI and agentic applications, right? Um obviously cloud is great, you know, pay as you go, infinite scale, global reach, um, you know, operational excellence, all of that is great, customers love it. But the one thing that kills them is the complexity, right? Um so many choices, which services do I need? How do I architect my app? And then even if I figure all that out, how do I actually deploy, make sure it's secure, scalable, cost efficient? So those were some of the challenges that we saw when we started DFANG a couple of years ago. Now, more recently, um you know, we're now in the agentic era, uh, coming from client server to SaaS to now agentic. And one of the trends we clearly start to see in the market is the deployment model for agentic apps is different from SaaS apps. And if I may just take a moment, explain what I mean by that. You know, in the SaaS world, the provider, the application provider, let's say a workday, uh, runs the application in their environment. And the customer just needs like a browser and a credit card to sign up. And all the complexity, the deployment, the operations are handled by that provider, workday in this case. But as we enter agentic era, and there are lots of agents being built by uh different startups all around the world, ISVs, when they go talk to their customers and say, Hey, I have this, let's say, fraud detection agent. And the customer says, Great, let me try it. And they do it like a pilot, and then they say, Like, I like it. I, you know, I would love to run this in production, but I'm not gonna send you my data, uh, you running as a SaaS, right? You have to bring this agent to where I have already consolidated my data from different applications, applied my compliance and security posture to it. Uh, this is where that agent needs to run. And this being my cloud account, whether that's AWS or Google or any other, right? So now the deployment model for agentic apps is not just run one instance, a multi-tenant instance in the cloud. It is run these agents. If you have 10 customers or 100 customers, you have to deploy them into each customer's environment. So that's making things even more challenging than it already was. Um, does that make sense?

SPEAKER_00 5:19

Absolutely. Thank you for demystifying um, you know, the differences between, you know, SaaS-based deployments and um, you know, uh in in customer's environment. Now, the thing about it is for a startup founder, as as much as you've been one yourself, you have much fewer resources and you need to grow in scale quickly. And, you know, this whole challenge of AI startup founders trying to get around the complexity of deploying in customer cloud environments. How do you think they can navigate such a complexity and how do you help them?

SPEAKER_01 6:01

Yeah, that's kind of exactly why we built Defang, right? So, our point of view is that developers, startup founders, they should focus on building their application. Uh, they have the domain expertise, they have the technical expertise to build their application. They should not have to struggle with all the complexities of deploying to various cloud environments, to various customer environments. So the approach we take is we let developers use an industry standard format called Docker Compose, which millions of developers already use, to define their application in what we think of as kind of a logical manner, right? Distinct from all the physical manifestation on each individual cloud. And then what DFIN does is do that translation. Um, so the same application, no changes, can be deployed to AWS, can be deployed to Google in different regions, in different customer accounts, optimized for different uh we call deployment modes. So whether it's production or development or somewhere like staging in between. Um, and all done correctly, all done according to the best practices of each cloud, uh, secure, scalable, cost efficient from day one, and can scale with you as you scale the business. So that's what DeFine does, like make deployment really, really simple. So startups can focus on building their applications.

SPEAKER_00 7:26

That's incredible. Um, you know, the world of cloud of late has you know gotten a lot of traction. Particularly, they've started discussing the concept of sovereign, with sovereign becoming sort of a little bit of this new buzzword. Perhaps to educate the audience and myself included, could you talk us through why sovereign is important?

SPEAKER_01 7:53

Yeah, and we see that conversation happening in a lot of different regions. We are based here in Vancouver, BC, so there's conversations in Canada, there's conversations in Europe, and pretty much every country is talking about how can we make sure that the data that you know companies here in this jurisdiction are building, that data stays in that jurisdiction, right? And some of them can use a local uh data center from one of the hyperscalers, and that's sufficient. But in other use cases, they may want a cloud that's actually owned and operated by a local entity, whether that's a consortium or a sort of partially government entity or whatever it might be. But the crucial challenge is not so much in building the hardware, like you know, you can build a data center, buy the chips, install the networking, storage, compute. But it's really like what is that software platform that application developers are going to target? And why are they going to learn a new and unique and niche cloud platform if that's only applicable in one small jurisdiction, right? That's where they struggle. And that's kind of where we are trying to help them is say, look, you can still have your developers, your ISVs, your application developers write to a common logical application framework. And then you can use a tool like Dfang to deploy it to the sovereign cloud as well, just like we do today for the hyperscalers. So that kind of gives you the best of both worlds. You can have the sort of locally owned platform, but developers don't have to customize their application to you know a very small platform potentially.

SPEAKER_00 9:34

Wow. Prakash, you have a knack for explaining very complex concepts in a deeply insightful and simple way. Thank you for such a deeply insightful conversation. And that's unfortunately all the time that we have for today. To our listeners, thank you for joining us. And don't forget to like, share, and follow our podcast. Our guest today was Prakash Sundaration of Defense Software Labs. And my name is Ed, your host of the AIP podcast. Cheers to a spectacular 2026 and wishing you all good health and much success. Thank you.

SPEAKER_01 10:15

Great, thank you.