The office's 22-year IT background didn't stay in the past — it became the foundation for its AI practice.
As clients and portfolio businesses started asking how to use large language models without sending sensitive data to a public cloud API, the same infrastructure discipline that built networks and data centres was pointed at a new problem: designing and standing up private, self-hosted AI environments. GPU servers are specified, procured and racked to match the model actually being run — not oversold on paper — and open-weight large language models are deployed and tuned entirely within infrastructure the client owns or controls. Nothing about a prompt, a document or a model's output has to leave that environment.
This matters most for organisations with genuine data-sovereignty requirements — legal, financial, healthcare and family-office contexts across Singapore and Malaysia, where sending proprietary or personal data to a third-party model provider isn't a real option. The office builds the alternative: an AI system that behaves like owned infrastructure, because it is.