H2O.ai opens Singapore lab for sovereign AI projects

H2O.ai has opened a forward-deployed AI lab in Singapore to support locally built artificial intelligence projects.

The move places senior engineers close to customers in sectors including finance, healthcare and public services, founder and Chief Executive Officer Sri Ambati said. He said the company chose Singapore because its tight labour market, limited land and AI-aware customer base create pressure to deploy working systems rather than produce demonstrations.

Ambati described the city-state as a demanding market where government agencies and private companies understand AI well enough to test products rigorously. That, he said, makes it a useful place to develop and refine systems in real operating environments.

The lab is part of H2O.ai's broader push around what Ambati called sovereign AI. In his view, that means a country should be able to diagnose, debug, improve and build its own core AI systems rather than rely solely on licensed tools developed elsewhere.

For Singapore, he said, that would require local data, local infrastructure and local talent. H2O.ai plans to develop that talent by placing Singapore-based engineers on live projects with experienced staff, including Kaggle Grandmasters, rather than relying only on courses or workshops.

Local focus

Several areas are under consideration for the Singapore operation, including tools to identify scams earlier, AI bots designed to engage suspected fraudsters and extract intelligence, robots that can assist elderly residents, and models aimed at predicting lightning strikes near high-rise buildings.

Ambati also pointed to women's health as a possible area for work with hospitals. He argued that some of the earliest practical gains from AI are likely to emerge in science and medicine, where shorter research cycles could bring discoveries sooner.

H2O.ai's delivery model centres on engineers working on site with customers. Ambati said physical proximity matters because teams inside banks or hospitals can identify operational gaps that would not appear in a polished demonstration.

"Because those limits do some of the work for us. When you don't have spare land or spare people, you end up leaning on AI and robotics just to keep productivity moving, so the pressure to get it right is higher here than in most places we operate. And the customers are demanding. Government agencies, private companies, they're already AI-literate, so our products get pushed harder in Singapore than they would somewhere more forgiving. I'd rather build against that kind of resistance than in a market that lets us coast," said Sri Ambati, Founder and Chief Executive Officer, H2O.ai.

He tied that thinking to a broader argument about national control over AI systems. "I'd compare it to a power grid, or water supply. You either own and control your core AI systems or you don't, and using something isn't the same as owning it. A country becomes AI-sovereign when it can diagnose a model, debug it, improve it, build on it, not just license someone else's and stop there. For Singapore, that means building up local data, local infrastructure and local talent, instead of importing all three off the shelf," said Ambati.

Skills model

Ambati said H2O.ai wants its Singapore hiring and training model to focus on direct exposure to production work. "We put local engineers on live projects next to our most senior people, some of whom are Kaggle Grandmasters. Not a course, not a workshop, an actual project with actual stakes. Sitting next to someone who's already solved the hard version of your problem teaches you faster than any curriculum does, and honestly Singapore starts ahead with her strong university ecosystem and community-centric culture, so we're not starting from zero," said Ambati.

On delivery, he said the lab is designed to help move projects from pilot stage into measured use.

"Forward deployment. Senior engineers sit physically close to the customer instead of running things remotely from some head office. It's a product decision, not just a staffing one. Spend enough time inside a bank or a hospital and you find the gaps a demo would never show you, and it's closing those gaps, not polishing the pitch deck, that actually matters. What we're after is getting past the demo and into something that produces a return you can measure. We call that the last mile," said Ambati.

The company's list of local use cases reflects issues specific to Singapore's urban density and demographics.

"A few things are on the list. On financial crime, the idea is to build models that catch scams earlier, and to go a step further with AI bots that pretend to be potential victims, tying up a scammer's time and pulling intelligence out of the conversation. We also want to look at robots that can safely help elderly residents, and at predicting lightning strikes near high-rise buildings, both shaped by how dense Singapore is and by its climate, so any model would need guardrails tuned to local conditions rather than a generic template pulled from somewhere else. In healthcare, women's health is underfunded, and that's an area where I'd like to move quickly with local hospitals. More broadly, I think the nearest payoff from AI shows up in science and medicine, where it can cut years off getting to a discovery," said Ambati.

Longer view

Ambati also linked the company's healthcare interest to a personal motivation. "My mother was diagnosed with cancer, and that's really where it started. I wanted AI pointed at the disease directly, and that turned into something bigger for me: making AI cheap enough, fast enough, easy enough that it becomes ordinary. Like water. If it costs less to run, more people get to experiment with it instead of a handful of well-funded labs deciding what gets built. That's not a slogan we put on a slide, either. It's the reason we've managed to pull in people to work on things like counting endangered species, alongside the paid healthcare work," said Ambati.

He also set out a broader view of how AI may develop over the rest of the decade, from smaller language models that can run on devices to robotics and scientific research.

"I think AI adds more than US$100 trillion to global GDP by 2030 and beyond, and I want human health and longevity to be somewhere in that story with H2O.ai's name on it. The shape of the technology itself is going to change too. Today's large models are dinosaurs, honestly, and better data, including synthetic data, should push things toward small language models that can run on a device rather than a data center. After that, physical AI, robots, is the next big shift, and I'd bet it moves faster and matters more than most people expect. Space exploration too," said Ambati.

Asked what success in Singapore would look like, he said: "Whether a small country with real constraints can become a maker of AI instead of just a customer for it. If that works, other countries chasing the same thing get a blueprint they can actually use, one built on co-creating with local partners, on returns you can measure, and on homegrown people, not just a licensed product shipped in from somewhere else," said Ambati.