categories

HOT TOPICS

Capital Efficient Strategy for Generative AI Startups: Aisera CEO Muddu Sudhakar (Part 2)

Posted on Tuesday, May 30th 2023

Sramana Mitra: Let’s get into some of those questions. You mentioned Azure. Now, generative AI from the Microsoft camp is available on top of Azure as a PaaS for companies like Aisera to tap into.

Muddu Sudhakar: We partnered with Microsoft. We use their Azure Open AI. Our algorithms and NLMs are built on Azure.

Sramana Mitra: All the computing is happening on the Azure platform?

Muddu Sudhakar: Exactly. It’s like build on Azure and execute on Azure Cloud for Microsoft Teams users.

Sramana Mitra: But the domain knowledge is coming from Aisers.

Muddu Sudhakar: The domain NLMs and the customer understanding are coming from us. Can I bring the network effects and learning from the other customers so that you also get the benefit. When you ask me a question about my WiFi or my network, I know what that means. I don’t have to learn from you.

Sramana Mitra: The IP on that learning is with you.

Muddu Sudhakar: All the algorithms are our IP.

Sramana Mitra: How mature is the Azure Open AI platform at this point?

Muddu Sudhakar: It’s mature enough for us to build. We have close to 100 million users on our platform. We build this platform on top of them. If their service goes down, our service goes down.

Sramana Mitra: What should be the entrepreneurship strategy of new entrepreneurs who want to build on top of the Azure Open AI stack?

Muddu Sudhakar: Two things. One thing that’s important is to leverage what exists. You don’t want to build everything by yourself. On top of that, if there are any platform services, use them too. You can also do AIaaS. Azure Open AI will let you build better algorithms and then you add your value-add.

Sramana Mitra: Are you following other companies who are trying to build platforms?

Muddu Sudhakar: To be honest with you, I have questions about those things. So much money has gone into proprietary NLMs. What happens to the cloud? What happened after 10 years? Only four clouds survived.

Sramana Mitra: There are a couple of factors going on in the market now. There are large players who already have large cloud infrastructure as a service businesses layering in generative AI. making AI available as a service is very easy for them to do.

Muddu Sudhakar: Today, I’m using Azure because we’re very happy with Azure Open AI.

Sramana Mitra: Generative AI is a massively infrastructure-heavy game.

Muddu Sudhakar: It is. The underlying infrastructure is now being optimized for that. These cloud providers will optimize it on their compute. Other proprietary NLM companies, in the long term, are going to lose. The investor who invested in them will make their portfolio companies run it on them. Not only are you losing money on them, but the portfolio companies will also lose. They’re creating a barter system. The money goes from company A to company B.

Sramana Mitra: I love what you’re saying. I’m looking at these fat startups that are trying to build their own proprietary NLMs, but compute is expensive. If they want to do their proprietary NLMs, the only strategy they have available is a PaaS with which they have to build business. PaaS is very difficult to build. We have Salesforce.com. A whole developer ecosystem is difficult to build.

If you’re an entrepreneur who’s not trying to build a highly-capitalized startup but trying to build a lean startup, that entrepreneur is going to build on Azure, Google, or Amazon who will give them a fast entry point.

This segment is part 2 in the series : Capital Efficient Strategy for Generative AI Startups: Aisera CEO Muddu Sudhakar
1 2 3 4

Hacker News
() Comments

Featured Videos