ABSORB PAIN.SHIP PRODUCT.
We’re standing up a forward deployed engineering crew at Kinetic IT: one Head of AI Engineering & Platforms, four Forward Deployed Engineers, zero layers in between. Each FDE embeds inside one enterprise account — on-site, next to the people who do the work — and owns its technical outcome end-to-end. Discovery to working software in days, not quarters.
THIS PAGE IS A FILTER, NOT A FUNNEL. IT IS WRITTEN TO MAKE THE RIGHT FOUR PEOPLE LEAN IN — AND EVERYONE ELSE CLOSE THE TAB, WITH OUR GENUINE GOOD WISHES. IF READING IT MAKES YOU TIRED, IT JUST SAVED US BOTH AN INTERVIEW.
The gap is the job.
Three facts make forward deployed engineering the correct structure right now.
The adoption gap is the opportunity.
AI capabilities are racing ahead of the world’s ability to use them. Models can do astonishing things, and yet inside most enterprises the day-to-day feels unchanged. AI does not deploy itself. Closing that gap takes human ingenuity, change management, and a real tolerance for pain.
There is no incumbent product.
In AI agents, nobody has the product yet — not us, not the competition. The market is a pile of heterogeneous segments that will each turn out to want different things. When there is no incumbent, product discovery is the entire game.
Discovery only happens inside.
Sales-led discovery talks to people from the outside and produces feature requests: what is easy to ask for, which is rarely what is valuable. The discovery that works happens sitting next to the planner and the dispatcher, living their day. Customers don’t know what they want. They want a partner who figures it out with them, inside their walls.
So we sell outcomes, not software. A contract with us is a commitment that a specific, measurable thing in the customer’s operation will be transformed. Contracts grow as outcomes compound. The engineer at the frontline owns the outcome end-to-end.
That engineer is you.
Forward deployed is a verb.
The name comes from the military. A forward deployed unit doesn’t sit at headquarters; it operates inside the environment where the problem is happening. Neither do we.
- Solutions engineering
- Consulting
- Professional services
- A ticket queue with travel attached
Operating. You sit at the customer site and fill the gap between what the platform does and what the customer needs — with working software, shipped into their environment, while they watch.
Most software companies are built inverted: high priests at HQ decide what to build, and field people implement it off a wiki page. We run it the other way. You figure out in the field what works. Then, together, we generalize backwards from the outcome. The whole company runs on that one loop.
Gravel road, paved road.
Everything in this org is in service of one loop. You build the specific. We extract the general. Repeat.
GRAVEL ROAD
You embed at your account, find the problem that actually matters, and build a working solution into their environment: fast and real, rough where rough is fine. The first version is allowed to be ugly. It is not allowed to be late, or wrong about the outcome.
PAVED ROAD
The Head owns the platform and the call on what generalizes. We don’t copy your solution into the platform — we solve the more general problem behind it, one level of abstraction up, inside the ontology, so it serves the next five accounts. Not just yours.
LEVERAGE
The paved version comes back to you as platform capability. Your next deployment of the same outcome takes a fraction of the time — and you spend the surplus signing up for harder problems, which feed the loop again.
ONE TURN OF THE LOOP, CONCRETELY
- TUE
On-site at account one. You sit with the dispatcher through the 6 a.m. shift and watch the first two hours of every day disappear into reconciling the schedule by hand.
- THU
A rough agent is live in their environment, doing the first pass while the dispatcher corrects it. First demo, four days in.
- +2 WKS
The FDE at account two describes the same shape in their async notes. The weekly session rules it a pattern.
- FRI
The two FDEs and the Head design the general version against both accounts at once. It lands in the ontology one level up from either customer’s vocabulary.
- NEXT
Account three deploys it in a week instead of six. That is the machine. Everything else on this page exists to keep it turning.
Five people are the product org.
One Head of AI Engineering & Platforms. Four Forward Deployed Engineers, each embedded in a different account. No PM layer, no solutions team, no product org down the hall. The crew of five is the product team — by design. Every hire is 20% of the product team and 100% of an account.
The account outcome.
What gets built at the customer, in what order, deployed how. Discovery, build, adoption, and the contract growth that follows. Nobody hands you a spec, because the customer doesn’t have one — they want you to figure it out. Loose, declarative direction in; a delivered outcome out.
The platform, the ontology, and the generalize-or-not call.
Made in the open, with reasons, every week. You will often disagree. Good — push back constructively and pointedly. That friction is the product process. When your gravel road is chosen for paving, you join the design alongside the FDEs whose accounts it will also serve.
THE HEARTBEAT
- THU NIGHT
Async notes in. What shipped, what’s blocking value, what the customer is really asking for underneath what they said, and the concrete ask: “this capability moves this contract from X to X+10%.”
- FRI
The weekly generalization session — the most important hour at the company. Four FDEs and the Head. Frontline learnings in, platform decisions out.
- +1 DAY
Decisions published. What generalizes, what stays bespoke, what is deferred and what evidence would change that. A decision that is not written down was not made.
- MONTHLY
Metrics review: platform share of each deployment, on-site weeks per FDE, contract trajectory per account. Same written decision note as everything else.
Outcome value per account.
Did the thing we promised actually change? Contract growth is the measurable proxy. A flat contract at a healthy account means we’ve stopped earning the right to harder problems.
Product leverage.
Time-to-deploy the same outcome at the next account, and the platform share of each deployment — rising, relentlessly. It is the single best indicator that gravel is becoming pavement.
All four seats are open and nothing above is running yet. No case studies, no reference accounts, no highlight reel. We’re publishing the operating model before the first hire, on purpose: written after the fact, it’s a description; written before, it’s a commitment. Every mechanism on this page is calendared and measured from the first week — and you can hold us to each one.
A week, for real.
On-site roughly a week at a time, multiple times a quarter per account, more during go-lives. This is what one of those weeks looks like.
Flight out. Go-live is Monday morning. You’re at the gate, not on a video call.
Go-live. Daily standups with the customer all week. Their last three big software projects failed, and they assume we’re the same. You earn trust by shipping, not by promising.
On the floor with the early shift. Living their day, learning their acronyms, finding the gap between what they asked for and what their morning actually looks like. Embedding, not visiting.
A hard meeting. Two hours of frustration, some of it unfair, all of it useful. The deliverable after a bad meeting is a better product, not a complaint.
Demo. Rough where rough is fine — live in their environment, on their data, in front of the people who’ll use it.
Async notes from the hotel: what shipped, what’s blocking value, what they asked for versus what they need, and the capability that moves this contract from X to X+10%.
Generalization session with the Head and the other three FDEs. Frontline learnings in, platform decisions out. Then home.
It is exhausting. That is exactly why it works — and why most people can’t follow you here. Pain is a moat. The relationships, the deployment speed, and ultimately the value all run through presence. If the travel is a dealbreaker, this is the wrong role — genuinely, no judgment.
Five situations. Two answers each.
No score. No percentage. The site doesn’t decide — you do. That’s the whole design.
Monday is go-live. The version that works is ugly: hard-coded edges, no tests worth bragging about. The version you’d be proud of needs three more weeks.
Seven traits. The first three are non-negotiable.
We hire on traits and treat backgrounds as evidence. Domain knowledge can be trained in weeks. The traits can’t be trained at all.
Curiosity that crosses domains
You walk into a freight yard, a hospital ward, a control room — somewhere you’ve never been — and within two weeks you see what nobody inside it can. Not curiosity about frameworks. Curiosity about forklifts.
A very high sense of urgency
Days-to-value is the operating spec. People with urgency talk in days. People without it talk in phases. You built the demo before the meeting where it was requested.
Customer obsession — genuine, not salesy
No one likes being sold. The pattern that works: engineers transparently excited about the customer’s problem, and the customer can feel it. The test is whether you talk about customers as people whose day you lived, or as logos you closed.
Resilience, translated into product
Hard meetings, skeptical rooms, weeks where everything you built gets ripped out. The real test is the transmutation: turning the hard week into Thursday-night notes, feature requests, platform improvements. Both kinds of people have war stories; only one has a changelog.
Appetite for ambiguity
Nothing is written. The right person hears “go make this account succeed” and comes back with a plan, a demo, and three things we didn’t know. The wrong person hears it and feels abandoned.
Prototyper, not cathedral-builder
Craftsmanship that needs perfect abstractions before shipping is a real and honorable skill — and it is not this job. The first version is allowed to be ugly. It is not allowed to be late, or wrong about the outcome.
The splinter in the mind
The single best predictor: something to prove — that the outcome should happen, and that you’re the one who can make it happen. The splinter is what brings you back Monday. Everyone great has one.
The ticket-taker · The ivory-tower craftsman · The careerist · The salesperson in an engineer’s body · The homebody · The pure consultant · The box-checker
People who need a spec. People who’d rather not travel or talk to customers. People optimizing for title. We mean this kindly, and we mean it: you’ll do great work somewhere — it just won’t be in one of these four seats.
This is the closest thing to founding a company while wielding someone else’s platform leverage. The failed startup is often a better signal than the comfortable big-co career — you’ve already proven you’ll run at a wall.
The physicist who writes production code. The grad-school dropout. The electrical engineer who taught themselves distributed systems. People who came to software from somewhere else usually have a reason to be here.
Especially the ones who sat in customer calls. You’ve already lived “no spec, no PM, ship it this week.”
Strong engineers fed up with low-impact work — two years on a feature nobody used. You light up when the outcome is physical: the shipment got picked up, the line didn’t stop, the three-day report takes three minutes.
What we don’t require: elite-company pedigree, deep ML research credentials, or ten years of experience. The technical bar is real — you can architect, build, and ship production AI systems alone, across the stack — but it’s a bar of speed and resourcefulness, not pedigree.
The summary test: could you have started a company instead? Not did you — could you. If the honest answer is yes, keep reading. We’re now the ones selling.
Founder training, said proudly.
An FDE spends years doing exactly what a founder does — find the real problem, build the product, deliver the outcome, grow the account — with platform leverage a solo founder would kill for, top cover, and someone else’s burn rate.
FDE
Own an account end-to-end. Build gravel roads that carry production traffic. Earn the right to the customer’s harder problems and grow the contract behind the outcomes. Build a resume nobody else can replicate.
PLATFORM & PRODUCT
The field is the only path into core product here — full stop. You don’t earn the credibility to make generalization calls without frontline scars. The platform team of year three is the FDE crew of year one.
POD LEAD
As we scale past four FDEs into pods and verticals, the people who ran the loop at one account run it across several. Four people today; the leadership bench of the whole org tomorrow.
FOUNDER-GRADE OPERATOR
Here — or at the company you’ll eventually start. Some of our best people will leave to found things. That’s the strongest proof the role works. We’d rather spend three years with someone who could, than thirty with someone who couldn’t.
The honest trade: the hardest, fastest, most exposed engineering job we have, in exchange for compressed judgment — product, engineering, and business at once — at a rate no other role builds. You’ll never wonder if your work mattered. The thing that took their team three days takes three minutes, and you’ll watch it happen in person, because you’ll be in the building.
What’s behind you while you do it.
Be honest about your alternatives. A comfortable senior role where your work dissolves into a roadmap. Founding your own thing, and spending year one earning the meetings, the security clearances, and the trust this role hands you on day one. Or skipping to the part founders actually want: a customer, a hard problem, and the authority to ship.
AU$200–250K package + 10% STI. Positioned above senior engineering bands — this role generates revenue, it doesn’t cost it.
Four. With the Head of AI Engineering & Platforms, you are the entire product org. 20% of the product team, 100% of an account.
Roughly a week at a time, on-site with your account, multiple times a quarter — more during go-lives. Sunday-night flights happen. Non-negotiable, and the reason everything works.
Zero. You work directly with the Head. No translation loss, no committee between you and the platform.
2,600+ client sites across Australia. A decade inside Qantas Group. BHP. Water Corporation. St John Ambulance. WA Police, Justice, Education, Communities, Landgate. Government, aviation, resources, utilities, emergency services — enterprises with real operations and real stakes, where Kinetic IT’s crew has spent twenty years earning the kind of trust most startups die waiting for. Distribution and security clearance you cannot buy, already paid for.
The first four accounts are being chosen from that base now. No case studies yet — that’s the point. What we’re selling is the model, and the chance to be one of the four people who prove it.
Five yeses, or close the tab.
Nobody is counting except you — which is exactly how the whole job works.
Send two things. No cover letter.
- 01
Whatever best proves you can build: GitHub, a product, a demo, a deployment story.
- 02
The story of an outcome you shipped against the odds. No spec, skeptical stakeholders, missing infrastructure, hostile timeline — and it worked anyway. What was the resistance, what did you build, what changed for the people who used it, and what did it cost you. Half a page. Specifics over polish. If nothing comes to mind, this isn’t your job yet — and that’s genuinely fine.
FOUR SEATS. THEN WE GO DO THINGS THAT DON’T SCALE, AT SCALE.