Claiming a successful response to the recently launched Insurance AI Cloud platform, UnifyAI, Sharad Sanghi, CEO and Co-Founder, Neysa, and Sandeep Khuperkar, CEO and Founder, Data Science Wizards, talked about India’s potential as an AI-builder and barriers holding it back.
How are the customers responding to UnifyAI?
Sanghi: We are getting a great response and there is a very good traction in Europe and US as well. This platform will be a game changer in terms of AI adoption for insurance. It will accelerate AI adoption for the industry and go in production in 3 to 4 weeks net. By March we will be having 300 plus insurance agents as an agentic approach in the platform. And on top of that there is a capability for insurance company to create their own Gen AI agent within 3 to 4 hours. And take it into the production.
What challenges have you faced while working on this platform?
Sanghi: I wouldn’t call it challenges but customers generally are not confident about AI adoption. In the boardroom there’s a discussion that AI needs to be adopted otherwise we’ll lose out. But possibly because of lack of awareness or immaturity in this cycle of AI, they end up spending a lot of money while doing something that does not result in good ROI. So there’s that.
How is India viewed in the AI-building ecosystem?
Sanghi: If you compare to the US, we are not building as much. For example, in the US you have several companies building really good foundational models and you have a lot of domain specific companies building models for domain. In India, very few companies build foundational models. What Data Science Wizard and Neysa do together is provide the platform for people to start building and you also provide a services layer. Neysa marries machine learning operations with the infrastructure and DSW marries their platform with customers and uses their platform to build actual end user use case. So, companies like ours are changing that. Of course, I would not say India is only a consumer market. There are companies like Sarvam that has built a foundational model for Indic languages.
Should India build its own LLM infrastructure?
Khuperkar: We should be first contributing in LLM coding first in the open source community. How many people are contributing? Just building is a different thing but first we have to contribute. Tomorrow, if 80 per cent is participating to a particular by India it’s basically an India-built LLM.
Your Insurance AI platform also looks into data security. How do you feel your technology will deal with new forms of technological concerns like quantum hacking?
Sanghi: So to make it quantum safe we are partnering with several companies that have got technology today that assures quantum. We have used standard RSA 1204 or 2048 encryption but if you use quantum computing, in the future, all the encryption techniques will fail. We are also partnering with a company in Pune to adopt their technology for making quantum safe. They will become like a vendor to us or a partner to us to provide that quantum safe.
What is your take on India’s draft data protection rules?
Sanghi: The first earlier draft actually had criminal liabilities in case of data breaches which I didn’t agree with because even the government would have been an optimal breacher. NIC data centres had breaches too. So, I think the current draft, there is only financial liability. From a data sovereignty perspective, there is scope for countries that collaborate with India, have a reciprocal relationship with India. We can move certain kinds of data to those countries. So in the earlier draft there were three kinds of data. There was general data, sensitive data and critical data.
Sensitive and critical data had to be in India. Now I believe there is a little bit more flexibility as long as there is reciprocity with the country. So by and large, it looks much better than the earlier draft.
Do you feel India should continue with its light touch approach for AI regulation?
Sanghi: I think AI principles and no regulation is the right approach. The last thing we want is to come up with rules that end up constraining a technology you don’t fully understand. The classical example of this is in the telecom sector when they came up with rules around VPNs, etc. Do we need regulation in AI? Absolutely. We have instances of deep fakes and cybercrimes through AI. But it has to be done carefully. It has happened that across the world where the government tried to catch the bad guy and ended-up preventing the good things the technology could do as a result. So I think the light touch approach is the right thing to do. You still need to understand the implications of what you’re trying to do.
So other than insurance, are there any other verticals that you guys are thinking of branching out into?
Khuperkar: The next vertical is banking and telcom. But currently we have production clients in retail in Germany and manufacturing. We have already seen traction in BFSI, healthcare, research institutes doing models for Indian languages as well.
You got good responses for Series A Funding. Any plans on that?
Sanghi: First seed round we got ₹20 million and then another round of ₹30 million. We will do further fund raising next fiscal because we want to expand beyond India.
Sanghi: We now want to launch inference-as-a-service. We are also working on cybersecurities on data posioning protections, LLM firewalls, etc. The big goal is going beyond India.
Khuperkar: We want to make UnifyAI like an operating system of AI, like Linux, so it is embedded to GPUs and closer to data. In 6-9 months, data centres and AI is going to become prominent. Because the cost of movement of data would be much more expensive than adopting AI.
Published on January 25, 2025
