We build and run software that schedules work crews, texts them their jobs, and handles what they get paid. It runs live every day for real businesses. If it breaks, real people don't get told where to go, and real money goes to the wrong place.
An AI writes most of the code. We need someone to decide what gets built, and to catch the AI when it gets it wrong -- because it does, and it always sounds confident.
What we pay
$1,000 a month to start.
We review it after your first few weeks, and again once you've shown what you can do.
The step up from this job is
Client Lead
-- you'd run a client on your own and be the person they talk to. It pays significantly more, and we fill it by promoting someone who's already doing this job well. That's where this role is meant to lead.
What you'll do
Take a vague complaint -- "the schedule is wrong sometimes" -- and work out what's actually happening before changing anything.
Write down what a new feature should do, and just as importantly what it should not do, before any code is written.
Use the AI to build it, then check its work properly. Read the code like someone is trying to slip something past you.
Test on real data with the feature switched off, then switch it on and watch. Nothing goes straight to texting a hundred people.
When something breaks: stop the damage, find everything affected, then fix the cause -- not just the one thing that got reported.
Write short, clear updates. Most days that's how your team knows what's going on.
Who does well here
You've built something real. One thing is enough. Something messy, that other people ended up depending on. We'd rather hear about that than a list of technologies.
You don't take "it works" for an answer. A passing test and a green dashboard don't settle it for you. You go and check.
You think about what happens when things go wrong -- bad input, the same job running twice, someone in a different timezone, a message that says "sent" but never arrived.
You're fine saying "I don't know yet, here's how I'd find out."
You don't need a degree, a particular language, or years of experience. You do need good judgment, and we test for that directly instead of guessing from your CV.
This isn't for you if you want tickets handed to you with the thinking already done, or if you'd ship the AI's first answer because it looked right.
How to apply
https://volt-os-production-4bc9.up.railway.app/apply
Fill in a short form, verify who you are with Telegram, and a 20-minute timed test starts straight away. Do it when you have twenty quiet minutes and decent internet -- you only get one go at it.
In the form, tell us about one thing you built that mattered. What was unclear at the start, what you got wrong, and how you found out. That's the part we read closely.
What happens after
Timed test -- arithmetic, patterns, following written instructions (20 min)
Written questions about real situations from the system (~2 hrs)
A repo with a bad change hidden in it, and some unfamiliar code to get into (~4-5 hrs)
A conversation -- real problems, out loud (~2 hrs)
Paid trial, real work, everything reviewed (2-4 weeks)
Steps 1-4 are unpaid. The trial is paid from your first day.
Most people don't get past step 3. That's expected -- the whole job is about holding a high bar, so we'd rather find out early than three months in.
If you need something changed to take this fairly -- extra time, a screen reader, a different format -- just tell us before you start. Asking costs you nothing and isn't part of the test.