AI Forward Deployed Engineer

the job:

You get dropped into someone else's company and you build inside it.

You will not get a ticket or a spec written by a product manager. You get a business with real customers, real money moving through it, and a bottleneck that has been there for years. Your job is to go in, find out how the work actually gets done, and build the system that changes it.

Then you stay. You run the check-in calls. You fix it when it breaks, and it will break at inconvenient times. You sit with their CEO and teach them how the thing you built works, until they know it well enough to argue with you about it.

You own the client. Infrastructure, build, deploy, calls, support, teaching. Nobody sits between you and them, and nobody is going to break the work into tasks for you.

It is a lot of surface area for one person. That is the job, and it is why we are slow about filling it.

what we believe:

These are not posters on a wall. They are what you get held to in week one.

98% working is 0% working. A client does not see \"mostly done.\" They see broken. Trust gets built or destroyed in the last 2%, which is also the least interesting part to work on. Do it anyway.

A check that cannot fail is not a check. If you say something works, say what you ran to prove it. A test suite that times out reports zero failures and exits clean. A type check that runs out of memory prints no errors. Both look green. We would rather read \"I could not verify this, here is why\" than a confident green that turns out to be nothing.

Write down what you got wrong. Every failure we diagnose gets recorded with its root cause and the rule that stops it from happening again. A mistake nobody writes down is one we pay for twice. This is the opposite of blame. Nobody here has to pretend they have never taken down production.

Give the knowledge away. Clients pay us partly to teach them. The instinct is to hold a little back so they still need you, and it is the wrong instinct. Every capability you hand someone shows them ten more they did not know existed. Teach all of it.

who this isn't for:

Be honest with yourself here. It saves us both a call.

• You need the work broken into tickets before you can start.
• Nothing you have built has been used by someone who could break it.
• A deploy fails on a Friday and you go quiet until Monday.
• You would rather be the smartest person in the room than the most useful one.
• You are good at the code and would prefer not to talk to the customer.

That last one rules out more people than the other four together. Half this job is talking to people who are not engineers.

what we do not care about:

Certifications. Which school you went to, or whether you went. Years of experience as a number. Whether your GitHub has a green streak.

We care what you have built and whether anyone used it.

stack:

Next.js, TypeScript in strict mode, Tailwind, Supabase with an isolated database per client, Railway, and some Python services around the edges.

Knowing this stack is worth very little to us. Picking up an unfamiliar one fast and shipping on it is worth a lot. We are not going to quiz you on framework trivia.

terms:

Vancouver. Salary, with equity on the table for the right person. We will talk numbers in the first conversation.

• One more thing, and we mean it. This is a job post, not a promise. If nobody clears the bar we hire nobody and the seat stays open. We would rather leave it empty than fill it badly.