Australian owned · Newcastle, NSW

AI on your infrastructure. Your data never leaves your network.

I'm Michael Jennings. I help Australian organisations deploy serious AI capability on their own hardware: sovereign, private, and independent of US cloud providers. Architecture, deployment and hands on support from an engineer who has built these platforms end to end.

Start a conversation How I work

No pitch. No obligation. A straight answer about whether this makes sense for your organisation.

1,000+

Enterprise users onboarded in nine months to a sovereign AI platform I built

20M

Hours of potential annual savings identified in a Generative AI study I ran for a global enterprise

$50K

Saved per year by simplifying a data platform architecture and retiring under-utilised hardware

14+

Years across engineering, data, AI and infrastructure in Australian enterprise and government

Career figures from roles in large enterprise and government organisations.

01 / The problem

Your people are already using AI. The question is where your data goes when they do.

Most Australian organisations haven't decided to adopt AI. Their staff decided for them. Documents get summarised, emails get drafted, contracts get "quickly checked" in a free browser tab. Every one of those prompts is company data leaving your network. If you haven't given your people a sanctioned way to use AI, they're using an unsanctioned one.

It leaves the country by default.Public AI tools process your prompts on overseas servers, under another country's laws and another company's terms of service. For organisations in health, finance, legal and government, that's not a detail. It's the whole question.

Your data can become their training data.Depending on the tool and the tier, what your staff type in may be retained and used to improve someone else's models. Most free tools default to yes. Most staff never check.

Shadow AI has no audit trail.When AI use happens in personal accounts on personal devices, you can't see what went out, prove what didn't, or answer the question when a client or regulator asks it.

Per-seat pricing punishes success.Cloud AI is billed per user, per month, forever. The better the rollout goes, the bigger the bill, and it never converts into an asset you own.

You're building on someone else's roadmap.Models get deprecated, prices change, terms get rewritten, and your only option is to accept it. Dependence on a single overseas provider is a business risk, not just an IT one.

None of this is an argument against AI. It's an argument about where it runs. Modern open weight models are genuinely capable, and they run happily on hardware you own, inside your network, under your rules. Waiting doesn't make the problem smaller. The unsanctioned usage is happening today. The organisations getting this right are giving their people something better than the browser tab: AI that's faster to access, tuned to their work, and built to keep company data inside the network.

02 / What I do

Sovereign AI deployment for organisations that can't afford to get it wrong

Most AI consultants will sell you a strategy deck. I'll give you a working deployment. Model selection, hardware sizing, infrastructure architecture and production-ready configuration, matched to your use case, your constraints and your team.

Architecture Consultation

A focused session on your specific environment: which models fit your workload, what quantisation and hardware they need, and what the full infrastructure stack should look like. You leave with a written recommendation you can act on, with me or without me.

→ written recommendation

Deployment Blueprint

Production-ready deployment configuration tailored to your hardware: inference engines, databases, reverse proxies, storage and security hardening. Documented so your ops team can stand it up and own it.

→ configs + runbook

Implementation Support

hands on help with questions, troubleshooting and tuning while your team deploys, until you're running independently. The goal is your independence, not a retainer.

→ hands on until independent

03 / How it works

From first conversation to running system

01
You talk. I listen.
A short conversation: no cost, no obligation. You tell me what you're trying to achieve and what your constraints are. If sovereign AI isn't the right answer for you, I'll tell you that too.
02
You get a plan.
An architecture session focused on your environment: model, hardware, stack and the costs to expect, delivered as a written recommendation your team and your board can actually read.
03
I help you build it.
Deployment blueprint plus hands on support while your team stands it up. I stay involved until you're running independently. Then I get out of the way.

Every engagement starts the same way: with step one.

Book the first conversation

Costs nothing. Commits you to nothing.

M. JENNINGS · NEWCASTLE, NSW

04 / Who I am

Michael Jennings

I'm a Technical Lead Engineer with fourteen years in large Australian enterprise and government organisations, across engineering, data, infrastructure and AI strategy. I've built sovereign AI platforms from scratch, inside a large enterprise and for my own company: fully on premises, no cloud dependencies, no data leaving the network. In one rollout, more than a thousand people were using a platform I built within nine months.

Data Craftsman is how I bring that experience to Australian organisations that need it: businesses in health, finance, legal, government, manufacturing and professional services, where data sovereignty isn't a preference. It's an obligation.

Enterprise and government background. Fourteen years inside large Australian organisations, including years building and leading technical teams in environments where security review is a way of life, not a checkbox.

Full-stack AI infrastructure. GPU hardware sizing, model deployment, RAG pipelines, vector databases, and production security hardening. I've built every layer myself, which is why my recommendations survive contact with reality.

Working deployments, not theory. Every platform I've built runs on premises today. My job with you is compressing years of my trial and error into weeks of yours.

What working with me looks like

You deal with me directly. No account managers, no handovers, no juniors.
If your problem doesn't need sovereign AI, I'll say so in the first conversation.
Everything I deliver is documented so your team owns it. I build your independence, not my retainer.
I keep client details confidential. It's the same standard I'm promising for your data.

If this sounds like your situation, get in touch. The first conversation costs nothing.

05 / Questions

Straight answers to the questions I hear most

What is sovereign AI?

Sovereign AI is artificial intelligence that runs entirely on infrastructure you own and control, inside your own network and your own country. Your data is never sent to an overseas provider, never processed under foreign law, and never used to train someone else's models. For Australian organisations with data sovereignty obligations, it's the difference between using AI and depending on it.

Is on premise AI as good as ChatGPT?

For most defined business workloads (document analysis, drafting, search over internal knowledge, summarisation), modern open weight models running on your own hardware are genuinely strong. Frontier cloud models still lead on some open ended tasks, which is why the real question isn't "which model is best" but "which model is right for your workload". Answering that question is exactly what an architecture consultation is for.

What hardware do you need to run AI on premise?

It depends on the models you run and how many people use them: anything from a single GPU server to a small cluster. Hardware sizing is part of my architecture consultation, so you know exactly what to buy before you buy it. Many organisations are surprised at how capable a modest, correctly specified setup can be.

How much does on premise AI cost compared to cloud AI?

On premise AI is mostly a one-off investment in hardware and engineering, while cloud AI is a per-user subscription that grows with your headcount and never ends. Which is cheaper depends on your scale and use case, which is why I scope expected costs in the architecture session before you commit to anything. You end the session knowing the numbers, not guessing at them.

Isn't an Australian cloud region already sovereign?

An Australian data centre region keeps your data onshore at rest, but the platform is still operated by a foreign provider under their terms, their pricing and their roadmap. True sovereignty means you could keep running if that relationship changed. For some organisations a hybrid approach is the right answer, and if that's you, I'll say so.

What happens to company data staff put into public AI tools?

It is transmitted to the provider's servers, usually overseas, and depending on the tool and account tier it may be retained and used to train future models. Free consumer tools typically offer the fewest protections, and staff use them because they're the easiest option available. The practical fix is giving your people a sanctioned alternative that's better than the browser tab.

Why work with an independent consultant instead of a large firm?

With me, the person who scopes your architecture is the same person who has built these platforms and the same person supporting your deployment. There's no handover to a junior bench. Everything I deliver is documented so your own team owns and runs it, which means you're never dependent on me. You're buying a working system and the independence to operate it, not a relationship with a firm.

06 / Get in touch

Let's talk about your AI deployment

Tell me what you're trying to achieve and I'll give you a straight answer about whether I can help. No pitch, no obligation, no follow up sequence. Just a conversation about whether sovereign AI makes sense for your organisation.

I reply within one business day
a short scoping conversation
no obligation either way

Prefer email? michael@mail.datacraftsman.com.au

Message me on LinkedIn

Fastest way to see who you're dealing with before you say a word.