Build AI systems that are still running in a year

Today this is one engineer and six systems in production. These are the roles we are opening as that stops being enough. Every band below is the real number, and the work is client work, so you will know within a month whether what you shipped survived contact with real data.

Remote across the EU and the US. Written by Lukas Ceponis, who will be on your first call.

Why this rather than a larger place

You will see the whole job

Scope, build, ship, and the month after. Nobody here hands you a ticket detached from the client who asked for it.

Prices are published, so scope is real

Every number we charge is on the website. That removes the negotiation that usually turns into unpaid scope creep on your week.

Small enough to fix things

If a process is stupid, say so on a Tuesday and it changes on Wednesday. There is no committee to route it through.

Production, not pilots

Six systems run today and people notice when they stop. That is a harder and more interesting problem than another proof of concept.

Open roles

Five of them, none filled. If two of these describe you, apply for the one you would rather do all day and say so in the email.

Engineering

Automation Engineer

$55,000 to $85,000 plus equity

Remote, EU or US · Full-time · 3+ years

The core of what we sell. You take a process somebody currently does by hand and put it into production so it survives real data and real users.

What you would do

  • Build and harden workflows in n8n, Make and Zapier, including the error handling, retries and alerting that separate a demo from a system.
  • Take rescue work: read somebody else's broken build, find what actually failed, and write the diagnosis before touching anything.
  • Own delivery on fixed-price builds from scope to handover, including the runbook the client keeps.
  • Watch the systems we run on care plans and fix what breaks inside the window we promised.

What we need to see

  • Production automation you can show us. A repository, an exported workflow, or a written account of something that broke and how you found it.
  • Comfort reading an execution log and reasoning about what the data looked like at each node.
  • Willingness to say a project should not be built when that is the honest answer.
Apply for Automation Engineer

Voice AI Engineer

$65,000 to $95,000 plus equity

Remote, EU or US · Full-time · 3+ years

Voice agents that answer a real phone line, hand off to a person when they should, and do not invent an answer when the tool call comes back empty.

What you would do

  • Build and run realtime voice agents on LiveKit and the OpenAI Realtime API, with SIP telephony behind them.
  • Chase latency to the percentile rather than the average, because the tail is what callers hear.
  • Write the evals. Every prompt change ships against a suite that places real calls and scores the transcript from the worker, not from our own recognition.
  • Own the numbers a client sees each month: answer rate, handoff rate, and cost per minute all in.

What we need to see

  • Something voice-related you have run in production, even a small one, and an honest account of what went wrong with it.
  • Understanding that telephony is where most of the failures live, and comfort with SIP, DTMF and carrier behaviour.
  • A view on when a voice agent is the wrong answer.
Apply for Voice AI Engineer

Delivery

Solutions Consultant

$50,000 to $75,000

Remote, US hours · Full-time · 3+ years

The person on the call when a US or UK owner describes what is broken. You scope it, price it from the published ladder, and say no when the fit is wrong.

What you would do

  • Run diagnostic and audit engagements: read the systems, interview the people who use them, and write the report that names what is worth automating.
  • Turn a conversation into a fixed scope with an acceptance test, so nobody argues about done later.
  • Hold the line on published pricing, and tell a prospect when a cheaper service on our own ladder fits them better.
  • Stay on the account after handover so the client keeps talking to somebody who remembers the build.

What we need to see

  • You have scoped technical work for non-technical buyers and been held to the estimate.
  • You can read a workflow well enough to know when an engineer is being optimistic.
  • Writing you are willing to show us. Most of this job is documents.
Apply for Solutions Consultant

Technical Writer

$40,000 to $60,000

Remote, EU or US · Full-time or part-time · 2+ years

Every engagement we finish produces a document somebody has to live with. Runbooks, handover packs, diagnostic reports, and the writing on this site.

What you would do

  • Write the runbook for each system we ship: what it does, what it costs to run, what a normal day looks like, and the four things most likely to break.
  • Turn a rescue diagnosis into something the client's own team can act on.
  • Write for the site, which means real traces and real numbers rather than another listicle about AI.
  • Keep the published price pages and the offer copy honest as the offers change.

What we need to see

  • Technical writing you did for a real product, ideally where you had to learn the system yourself first.
  • You can read code and workflow JSON well enough to describe them without an engineer translating.
  • A low tolerance for filler. We keep a written list of phrases we do not use.
Apply for Technical Writer

Growth

Growth Engineer

$50,000 to $80,000 plus equity

Remote, EU or US · Full-time · 3+ years

Half marketing, half engineering. You build the tools and content that bring the right buyers here, and you measure whether they worked.

What you would do

  • Build free tools worth linking to, such as cost calculators and diagnostic utilities, and ship them properly rather than as landing pages.
  • Own the technical side of search and the AI answer engines, and be honest with us about which of the two is actually delivering.
  • Run the small paid experiments we can afford, and kill them on evidence rather than on hope.
  • Instrument the funnel end to end so we know which channel produced each qualified call.

What we need to see

  • You have shipped something that acquired users, and you can say what the numbers were.
  • Comfort with the fact that most marketing claims in this industry are vendor-published and unverified.
  • You would rather run one experiment properly than five badly.
Apply for Growth Engineer

How hiring goes

Three steps, no panel rounds, no take-home you do for free.

  1. 1

    A call, thirty minutes

    You talk to the person you would work with. We tell you what the work actually is, including the parts that are dull.

  2. 2

    A paid task

    Four to six hours of real work at your rate, on something close to what the job involves. We pay whether or not we go ahead.

  3. 3

    A decision in a week

    Yes or no within five business days of the task, with the reasoning either way. Nobody waits in silence.

What comes with it

  • The published salary band, paid in your currency where we can.
  • Equity on the engineering and growth roles, with the terms written down before you sign.
  • A budget for whatever AI tooling makes you faster, without an approval chain.
  • Remote, with the working hours the role names and no expectation beyond them.
  • Your name on the work, in public, on the case studies you deliver.

Nothing here fits

Send the work anyway. A repository, an exported workflow, a write-up of something that broke in production and how you found it. We keep a short list of people to call when a project needs a second pair of hands, and that list is how most of these roles will get filled.

hello@lurtoagency.com

We read every application ourselves. If you do not hear back within ten business days, write again, because it means we lost it rather than that we decided.

You are here to buy, not to apply

Every price is published and the scope is fixed before any invoice. Start with what is broken or what should stop being manual.

See the pricesBook a call