Home/About
Northlark exists because the distance between an AI demo and a system a business runs on is mostly work nobody enjoys selling — and almost nobody wants to do.
Why this exists
AI capability isn't the bottleneck any more. What stops a normal company getting value out of it is everything around the model: a process nobody has mapped, data spread across four systems, no clear owner, and a rollout that assumed people would change how they work because someone sent a memo.
None of that is solved by a better model, and none of it is solved by a strategy deck. It gets solved by someone sitting with the team, doing the integration, writing the tests, and staying until the thing is genuinely in use. It's unglamorous, it's specific to your business, and it can't be done in one shot from the outside.
That's the work I do, and I've deliberately kept this small so I can do it properly: few clients at a time, fixed prices, and a handover that leaves you independent. I'd rather have a client who no longer needs me and refers me on than a retainer that has quietly outlived its usefulness.
And to be straightforward: this is a new practice, and I'm early on my own book of work. What I offer in the meantime is a structure where the risk sits with me — a small fixed fee to start, fixed-price builds, and everything I produce transferring to you at every stage.
Who you're hiring
Founder — Northlark AI
Eight years in software, most of it spent building the systems businesses actually run on: a loans management platform for a financial services firm, control software for industrial 3D printers, education software used across K–12 schools, and a mobile app that put floorplans and material tracking into the hands of construction crews on site.
That range is the point. AI projects rarely fail on the model — they fail where it has to meet a real CRM, a real permissions model and a real team's habits. I've spent my career on that side of the work: React and Angular at the front, C# and Python behind, and the integration layer in between.
More recently I've been building AI systems held to the same standard. One uses Auth0 Fine-Grained Authorisation to scope a RAG assistant strictly to what each role is permitted to see, with third-party keys held server-side rather than shipped to the browser — the difference between a chatbot demo and something you can put in front of a business. Another is an MCP server that turns a Notion board into a live research desk driven by a local LLM, so nothing leaves the machine.
I hold a Master of Information Technology from Monash and a Bachelor of Accounting. The second one is why I read the business process before I read the codebase.
Fit
Half an hour is enough for both of us to know. If it isn't me, I'll say so on the call.