H2O.ai is expanding its presence in Singapore with a forward-deployed AI lab focused on applying artificial intelligence to locally relevant problems, from financial crime and healthcare to robotics and climate-related prediction. CEO and founder Sri Ambati says the strategy is designed not only to deploy AI for customers, but to build local engineering expertise and infrastructure around what H2O.ai describes as sovereign AI.
H2O.ai is taking a localized approach to enterprise artificial intelligence with a forward-deployed AI lab in Singapore, putting engineers closer to customers and positioning the country as a testing ground for sovereign AI, AI infrastructure and emerging physical AI applications.
In an interview with Kestone Global, H2O.ai founder and CEO Sri Ambati argued that Singapore’s constraints—including limited land, an aging population and a relatively small workforce—create practical incentives for organizations to use AI and robotics to maintain productivity.
The company’s strategy goes beyond selling AI software into the market. H2O.ai is placing senior engineers alongside customers and local teams to develop and deploy systems in real-world environments. Ambati describes this approach as “forward deployment,” with the objective of moving AI projects beyond demonstrations and into measurable production outcomes.
That model reflects a broader change in enterprise AI development. As organizations move from experimentation to deployment, the technical challenge increasingly involves integrating AI into existing workflows, data environments and operational systems rather than simply demonstrating model capabilities.
Sovereign AI moves beyond model ownership
A central theme of H2O.ai’s Singapore strategy is sovereign AI.
Ambati describes sovereign AI in terms of a country’s ability to understand, diagnose, modify and build on its core AI systems rather than simply licensing technology developed elsewhere. Under that definition, sovereignty encompasses local data, infrastructure and technical talent alongside access to AI models.
The concept has become increasingly important as governments evaluate their dependence on foreign cloud infrastructure, foundation models and AI development platforms.
Singapore has already established a national AI strategy and invested in AI infrastructure and talent development. H2O.ai’s approach adds a private-sector implementation layer by placing engineering resources directly into local projects.
For enterprise AI, the model can also address a practical issue: highly regulated organizations often need AI systems adapted to their specific data, processes and compliance requirements.
Forward deployment targets the last mile
H2O.ai’s forward-deployment model puts senior engineers physically closer to customers rather than relying exclusively on remote product development.
Ambati argues that this proximity reveals operational problems that are difficult to identify during demonstrations. The goal is to understand how AI performs within an actual bank, hospital or other organization and then adapt the technology to those conditions.
This is particularly relevant to machine learning infrastructure, where the difference between a successful prototype and a production system can involve data quality, monitoring, integration, model governance and workflow changes.
H2O.ai has historically positioned its AI and machine learning platform around enterprise model development and deployment. The Singapore initiative extends that philosophy into a more embedded delivery model, where engineers participate directly in solving customer problems.
Singapore becomes a testbed for physical AI
Several of the projects discussed by Ambati extend beyond conventional enterprise software.
H2O.ai is exploring financial-crime models intended to detect scams earlier, as well as AI systems that could interact with suspected scammers to consume their time and collect intelligence. The company is also looking at robotics for elderly assistance and models designed to predict lightning strikes around high-rise buildings.
These applications illustrate the growing convergence between AI software and physical environments.
Robotics, in particular, is becoming an increasingly important part of the AI technology stack as improvements in computer vision, machine learning, simulation and multimodal models make it possible to build systems that interact with the physical world.
The proposed Singapore applications also highlight why local data and operating conditions matter. A model designed for elderly-care robotics or weather-related prediction needs to account for the environment in which it operates rather than simply reproduce a generalized model developed elsewhere.
Healthcare remains a major target
Healthcare is another area Ambati identifies for the Singapore lab, including women’s health and research applications.
He also argues that science and medicine could be among the areas where AI produces some of its most significant long-term benefits by accelerating research and discovery.
Those ambitions align with a broader expansion of AI beyond business automation into scientific computing, drug discovery, medical research and clinical applications. However, healthcare deployments carry substantially higher requirements for validation, privacy, safety and regulatory oversight than many conventional enterprise AI applications.
H2O.ai’s stated objective is therefore less about deploying a single general-purpose model and more about applying AI systems to specific problems where local expertise, data and engineering constraints can be incorporated into the solution.
Smaller AI models and physical intelligence
Ambati also expects the AI architecture itself to change. He argues that today’s large models will increasingly be complemented or replaced in some applications by smaller language models that can operate closer to the device.
That trajectory would have implications for AI infrastructure. Smaller models can potentially reduce inference requirements and enable more processing at the edge, although the appropriate architecture depends heavily on the complexity of the task, available compute and accuracy requirements.
The next stage, in Ambati’s view, is physical AI: robots and autonomous systems capable of operating in the real world. He also points to space exploration as another area where AI could expand.
These remain forward-looking views rather than established market outcomes. The near-term test for H2O.ai’s Singapore strategy is more concrete: whether embedding engineers with local organizations can turn AI projects into measurable production systems while simultaneously developing local technical capabilities.
For Singapore, the initiative represents a different proposition from simply becoming a consumer of imported AI technology. For H2O.ai, it provides a real-world environment in which enterprise AI, machine learning infrastructure, robotics and sovereign technology strategies can be developed together.
Market Landscape
Sovereign AI is becoming an important consideration for governments and enterprises seeking greater control over data, models, infrastructure and AI expertise. Singapore has invested in national AI capabilities while also positioning itself as a regional technology hub.
At the enterprise level, AI deployment is moving toward more specialized systems that combine proprietary data, domain-specific models and existing business infrastructure. This increases demand for engineering expertise capable of moving models from development environments into production.
The emergence of physical AI adds another layer. Robotics and edge AI require models and infrastructure that can operate under real-world constraints involving latency, connectivity, safety and local environmental conditions.
H2O.ai’s Singapore strategy brings these trends together through forward deployment, local talent development and projects designed around regional requirements.
Top Insights
- H2O.ai is using a forward-deployed engineering model to bring senior AI expertise closer to enterprise customers and production environments.
- The Singapore lab links sovereign AI with local data, infrastructure and talent rather than treating AI sovereignty as model ownership alone.
- Proposed projects span financial crime, healthcare, elderly-care robotics and environmental prediction, extending AI into physical-world applications.
- Ambati expects smaller language models and edge AI to complement today’s large models as inference increasingly moves closer to devices.
- The Singapore initiative positions local engineering capability as part of AI deployment rather than relying exclusively on imported technology and expertise.
Power Tomorrow’s Intelligence — Build It with TechEdgeAI
