AI, APIs, and the messy middle of getting either into production.

I’m Duo (Neo) J. — currently an AI Solutions Engineer at Kong, working with enterprises on how AI actually reaches production: through gateways, policies, identity and observability rather than through demos.

In practice that means designing the layer between an organisation’s applications and the models they call. Multi-provider routing and failover. Semantic caching to cut latency and token spend. Token-based rate limiting and cost attribution per team or tenant. Prompt guardrails and data-loss controls on the way in and out. And increasingly, taking the APIs a company already has and exposing them to agents as governed tools over MCP — with real authentication, scoping, and an audit trail behind every call.

I came to AI the long way around: over a decade in data centres, cloud architecture and DevOps, then years as a pre-sales architect translating dense technical designs into outcomes a business could act on. That background is why I care less about which model is winning this quarter and more about the platform that has to carry all of them. In most organisations, the AI programme and the API programme turn out to be the same programme.

I learn in public, write at blog.duoj.au, and I’m most useful to teams who are part of the way through a transformation and need someone to help them name what’s actually in their way.