// about
A small engineering team who'd rather ship than pitch.
DevCrunch is a small team of engineers. No account managers, no sales team. If you work with us, you work with the people writing the code — the same people who built TroikaBot, Handman and DietApp end to end.
How we actually work
You talk to the engineer, not about them
There is one Slack channel per project and the people in it are the people building your system. No status meetings that exist to be reported upward, no requirements translated twice before they reach a keyboard. If you ask why a query is slow, you get an answer from whoever wrote it, usually the same day.
Two-week cycles, demoed to you at the end of each
Every fortnight ends with something running that you can click on. Not a slide about progress — the actual thing, on a staging environment you have access to. If a cycle produced nothing demonstrable, that's a conversation we'd rather have at week two than week twelve.
The repository is yours from day one
We commit to your GitHub organisation, not ours. There's no handover event at the end because there's nothing to hand over — you've had it the whole time. Infrastructure lives in your cloud accounts under your billing. We've had clients take a project in-house halfway through and that went fine, which is the point.
We measure before we optimise, and we say when we can't
Most AI work arrives with a claim attached: it'll be faster, it'll be more accurate, it'll handle the overflow. We build the measurement first so that claim can be checked. Sometimes the measurement says the idea doesn't work. We'd rather tell you that in week three than build something for four months that nobody trusts.
We turn down work that isn't ours
We don't do design systems, we don't do mobile apps, and we don't take on projects where the interesting part is a procurement process. Six areas, listed on the expertise page, and we say no to the rest — which is most of the reason the ones we take on go well.
How we're set up
We work remotely and schedule around our clients rather than the other way round. Our calling hours are chosen to overlap with European afternoons and with the US East Coast first thing, so a call lands in a normal working hour on your side rather than ours. Most of our work is for teams in Europe and North America.
We're deliberately not trying to become a fifty-person agency. We like being the size where everyone knows what's in every repository.
What we won't do
-
Bill you for a demo that can't go to production
If the approach won't survive real data and real load, we'll say so during discovery rather than build a proof of concept that impresses a board and then dies.
-
Put a junior on your project and call them senior
We do hire and train junior engineers. They pair with someone senior on internal work and on our own products, not silently on your build.
-
Hold your infrastructure hostage
Your cloud, your repos, your API keys, your billing. We ask for access, not ownership, and you can revoke it in an afternoon.
-
Add an LLM to something that needs a SQL query
A surprising share of "AI projects" are a reporting problem wearing a costume. We'll tell you when yours is, even though the SQL job is smaller.
// start here
Come and check whether any of that is true.
The intake form takes two minutes, then you pick a time. You'll be on the call with the engineer who would lead your build.
Book a call