Engaging with govt on AI, see big opportunities: Krutrim's Ravi jain

The vision is to enable the developers to do cutting-edge innovation. These platforms or tools that developers need are computing and services and that is what we have opened up

Krutrim is exploring the possibility of working with the central government to boost its artificial intelligence (AI) vision, says Ravi Jain, head of strategy. In an interview with Peerzada Abrar in Bengaluru, Jain adds that working with the Centre is a huge opportunity as it gives scale to the company. Jain says the company, focused on developing full-stack AI capabilities, has opened up its Cloud platform to enterprises, researchers and developers. Edited excerpts:

What are the innovation bets that you are making for the future?

The vision is to enable developers to do cutting-edge innovation. The platforms or tools that developers need are computing and services and that is what we have opened up. For example, the LLM (large language model) that we have, to create our assistant product is now available to developers to create their own products.

They can do this in whichever domain they want to and the infrastructure they need to fine-tune, pre-train and run other processes would be available on our Cloud. Whatever we create should be available to the developers to harvest. For instance, we have launched GPU-as-a-service, model-as-a-service and location services. Krutrim Assistant app is now available on Android and an iOS app would be coming soon.

What are the problems that you aim to solve?
 

The app that you have seen in the Web form, is much easier to use as a standalone native app. Going forward, we will do a lot of work on product features. Our thought process is that we may not be able to build all forms of applications for different domains. That's the reason why we want to give these elements or tools out to developers. They can use them and apply their innovation and create their models. For example, a healthcare company wants to use the patient interaction tool. It can use a model and create an Indian language patient interaction tool, which will help it provide services to Tier-III towns and rural areas.

How do you plan to generate revenue for the company through the products and services you are building?

There are many layers. There is revenue if people (enterprises) consume model-as-a-service and GPU-as-a-service as well as the general computing Cloud that we are building. These are massive opportunities. If you look at the comparable companies in the West, they are huge from a revenue and impact perspective.

How do you see the opportunity to partner with the government in AI?

We are engaging with it in different capacities to understand how the partnership would be meaningful. When there is anything concrete, the government would talk about it. Some aspects of the (government sector) are big opportunities for us. For example, government e-services use a lot of consumer interactions. There are potential applications for that. We don’t have anything to talk about at the moment. But we are looking at the government sector because that gives us scale.

Once a firm creates a model, some competitors may build better models. What is your strategy to tackle that?

It is a mix of both (improving models and leapfrogging). Let’s say there is a developer who has a preference for an open source model, available from a firm from any part of the world. We want to give that choice to the developer and don’t want to say that you should use only our model. In terms of Cloud computing, we are thinking about leapfrogging. We are not using the legacy of 10-15 years of Cloud technologies. We are using the new-age technologies to build Cloud from scratch. We have that advantage.

How do you aim to compete with tech giants such as Google, Microsoft, and OpenAI, which have dominated the AI space?

Our full-stack approach is going to make a difference. How you create models and services specific to India is also going to make a big difference. For example, for our location services, the amount of data and freshness that we have is real time. When you use location services on Krutrim Cloud, you want to do route planning.

If the road is obstructed, the real-time information we get is much richer because of all the assets we have on the roads. You can do very efficient re-routing using our location services versus others. There is India-specific data and our full-stack approach reduces the cost. India is one of the 180 countries where these (large tech companies) operate and their approach would be different. For us, it is front and centre.

Krutrim has plans to build its own silicon and design chips as well as data centres. How are those plans shaping up?

We are working on AI infrastructure to develop an indigenous data centre and, eventually, server computing, edge computing, and supercomputers. These are based on the latest technology nodes and packaging. It would take us some time to build the silicon which we would apply in our system as we scale up the Cloud. Meanwhile, we would partner with anyone who would provide us the infrastructure. Today, we need to provide that infrastructure to the developers.

We have partnered with Nvidia for this. Ultimately, we are going to provide our own chips. Even then, we would have our partnerships as each silicon is good at certain kinds of workloads. We are working with multiple partners, including startups, which are creating such technologies. However, getting access to such technologies may become difficult in the future due to the geopolitical situation. You can see the Chips Act in the US and its impact on China. We need to build indigenous technologies. We would also set up data centres across the country.