Prepared with AI assistance from the linked primary sources. Company and government claims are attributed to their issuers. How we use AI.

AI infrastructure depends on people who can keep a service useful after its initial demonstration. That includes data engineers, platform operators, application developers and domain specialists who can recognise a bad result. Without those capabilities, additional computing capacity may increase the number of experiments without increasing the number of dependable systems.

AI Singapore’s AI Apprenticeship Programme currently describes six- or nine-month routes. Both begin with three months of deep-skilling, followed by project work with experienced engineers. Its curriculum includes areas such as machine learning, large language models, MLOps and governance. AI Apprenticeship Programme description.

Practical experience fills a different gap

A course can introduce an algorithm or a development tool. A project exposes the constraints that determine whether a system survives contact with users: inconsistent records, changing requirements, limited capacity and ambiguous failures. Learning to investigate those problems is a capability distinct from reproducing a successful demonstration.

Employers also need to distinguish between building and operating skills. A prototype team may be comfortable changing a model quickly. A production team must consider compatibility, monitoring, incident response and the effect of a change on users. One person may perform several roles in a small company, but the responsibilities still need to be explicit.

Google’s September 2026 Singapore Engineering Center announcement included forward-deployed engineering and work across cloud infrastructure and enterprise AI. That provides a concrete example of a company connecting product engineering with customer implementation. It does not, by itself, quantify a regional talent shortage or establish training outcomes. Google’s engineering centre announcement.

Shared facilities need shared expertise

Research infrastructure has a similar requirement. NSCC’s ASPIRE 2B launch emphasised user enablement alongside computing capacity. The pairing recognises that researchers need practical support to use sophisticated systems effectively. NSCC’s ASPIRE 2B launch.

For a shared facility, an experienced support team can help users diagnose slow jobs, choose suitable tools and avoid repeating mistakes. The wider value comes when that knowledge is documented and reused. Funding a system without maintaining the supporting expertise risks making access difficult for precisely the institutions that most need assistance.

Measure capability after training

Training programmes should be evaluated through what participants can subsequently do. Useful evidence includes completing a deployment, explaining evaluation results, maintaining a service and responding to an incident. Attendance and certificates describe participation; they do not establish that an organisation can operate an important workflow.

Companies can make this more concrete by connecting training to a defined project with an owner, a quality threshold and a maintenance plan. They should also preserve time for experienced staff to mentor others. Otherwise, scarce experts can become permanent bottlenecks rather than a source of broader organisational capability.

Treating talent as part of infrastructure changes the investment question. The issue becomes whether the organisation has the combined machines, tools and expertise to deliver a service. That is a more useful measure of readiness than either the hardware budget or the number of people who have attended an AI course.

Sources

See something we should correct? Contact the newsroom.