LLM systems
that survive contact with users.
Designing the architecture around a model — prompts, retrieval, tools, fallbacks, evals — so it behaves the same on Tuesday as it did in the demo.
APPLIED AI·
LLM SYSTEMS·
AOTEAROA
A one-person studio run by Eaden McKee — for businesses who want to do something interesting with large language models, and a sharp pair of eyes on what's actually worth building.
OperatorEaden McKee
DisciplineApplied AI · LLM Engineering
LocationTāmaki Makaurau · Auckland
SignalLive
01OperatorWho you'd actually be emailing
The operator behind the signal.
Hi — I'm Eaden. I've spent the last few years going deep on large language models: how to design them into products, how to make them reliable, and how to tell the genuinely useful from the hype.
I currently work as an AI specialist at a top-10 New Zealand company. That role is my day job, and it's where my contracted hours live.
Superlinear™ is the studio name I've used for years for everything outside that — writing, prototypes, public talks, and chats with people trying to figure out where AI fits in their world.
I have an unreasonable patience for explaining things from first principles, and a fairly low tolerance for hype dressed up as strategy.
02PracticeWhat I actually know about
Applied AI, not theatre.
Six areas where I can almost always shorten your path — whether you ship the work or just walk away with a clearer map of the territory.
LLM systems
AI strategy
without the snake oil.
A short, honest conversation about where AI will actually move the needle for your business — and where it absolutely won't. Usually saves more than it costs.
Prototype
to production.
The journey from "this notebook is amazing" to "this is in front of customers at 3am" — observability, guardrails, latency, the things that bite.
Retrieval
that actually retrieves.
Most "RAG" projects fail at chunking, scoring, or ranking — not at the model. I can usually spot the broken link in ten minutes of conversation.
Eval frameworks
that tell the truth.
If you don't have evals, you don't have a product. Practical, cheap, fast feedback loops that let you ship changes without crossing your fingers.
Second opinions
and translation.
You've got a vendor pitch, a contractor's plan, or an internal proposal. I'll read it and tell you, in plain language, what you're actually being sold.
01
/
06
LLM SYSTEMS
03AudienceWho this is for
Built for any size of curious.
Size doesn't change whether AI is interesting for you — it only changes the shape of the conversation. Three rough flavours:
/ Small · 1 – 20
Founders & tiny teams
Figuring out if there's a real product in the AI corner of your idea, or whether the obvious thing has already been built. Usually a one-hour chat is enough.
/ Medium · 20 – 500
Scaling companies
You've shipped a couple of LLM features, you're not sure they're good, and you're not sure what to ship next. I can read the room and point at the highest leverage move.
/ Large · 500+
Enterprises & orgs
Internal teams, vendors circling, exec pressure to "do something with AI." A clear-eyed outside view — and someone happy to push back politely.
04AvailabilityRead this before emailing
PLEASE READ
— Important note —
I am not available for paid work, contracts, or consulting engagements. I am very available for a conversation.
I'm currently employed as an AI specialist at a top-10 New Zealand company, which is a role I love and intend to keep. That means I can't take on client projects, side contracts, or retainers through Superlinear™.
What I can do: get on a call, read your pitch, look at your prototype, argue with your vendor's deck — and if you need real hands on the keyboard, point you at someone in the NZ / Aus / global AI community who can actually do it. Often that's the more useful outcome anyway.
It's free. It's curiosity-driven.
— Send a signal —
eaden@superlinear.nz
Free. Always.
Reply within ~48h
No sales funnel
One human · NZ time
— Or save the card —