Not the right fit? Refer someone. If we hire them we’ll send you $500 — just make sure they name you on the application.
Open role · AI-Native DeveloperOpen role

Applications open · not actively hiring

We’re not actively hiring for this role right now.

We’re mid-interview on the current round, so this seat is close to decided. It isn’t closed, though. Leave your details and we’ll consider you for this role and for what opens next. No exercise and no timed stage — if there’s a fit, we reach out.

For AI-native devs & AI automation engineers

Work on AI systems already running for 7 and 8-figure businesses.

We do the marketing and the sales.

You join the build team.

Join the talent poolWatch the 1-minute video ↓

1-minute video with Ruben Davoli, Founder & CEO.

Our missionA two-person team, on purpose — until now.

BeaverMind is the white-glove AI growth partner for coaching and consulting businesses. We're operators first — we scaled a coaching company to $3M before building a single agent — so every system we ship starts from the revenue problem, not the tech. One rule runs everything: if a build won't produce clear ROI in revenue gained or time saved, we don't ship it.

Ruben Davoli

Ruben Davoli

Founder & CEO

Scaled an education company from $0 to $3M with 15,000+ clients — by building the systems that ran it without him.

After years of being let down by agencies, one pattern kept repeating: the priority was protecting the retainer, not the client's results. BeaverMind exists to break that model. Owns marketing, sales, and strategic vision.

Luke Cala

Luke Cala

Head of AI Solutions

The technical architect behind every system BeaverMind ships — AI voice agents, automation workflows, custom applications, and the full AI infrastructure stack.

Placed 2nd at Liam Ottley's AI hackathon, one of the most competitive AI builder programs running. A background in paid media means every build serves revenue outcomes, not engineering for its own sake.

Why hire now: demand has outgrown four hands. The client queue and the new builds compete for the same hours, and we won't let quality slip to grow. You come in as the third person — the day-to-day engine of delivery, working directly with Luke.

The goal isn't to be an AI agency. It's to be the AI growth partner for seven and eight-figure coaching and consulting businesses. You'd be joining the team that builds that.

Already runningThe systems you could be working on.

Everything we've built, are building, or have queued next. You'd start with one lane — Client OS — and grow from there; nobody owns all of this at once.

  • Client OS — our internal operations platform: client events, portals, payouts, call intelligence. The first thing you'd work on, and the base of your paid trial.
  • A certification program platform for an eight-figure coaching business — curriculum, assessments, AI-assisted grading, built as a 12-month AI partnership.
  • An AI intelligence layer for the same company — one knowledge base unifying their method, content, and delivery, feeding every agent they run.
  • Outbound and inbound voice agents on live phone lines — calling, qualifying, and booking leads end to end. One reactivated a dead list into $6.8K in 15 days.
  • A call-QC agent — reviews and scores every coaching and sales call against the client's own method, and coaches the team back. 100% review instead of a sample.
  • An AI Q&A clone of the coach — their expertise answering clients 24/7 inside the client portal.
  • An AI copilot for an advisory firm — ingests years of sales communications and surfaces the playbook patterns their team actually closed with.
The offerOne seat, one bar.

The hard part of this industry isn't the build. It's finding the clients, selling the work, and holding the relationship. We've done that — you own your piece of the build end to end and get paid for solving it. No agency to run, and no building alone either.

One thing we won't pretend: this seat needs someone who is already good. We're hiring to raise the bar, not to fill a chair — someone who pushes back, proposes, and shapes how we build. What you get in return is what good engineers rarely get: real clients, real production, real ownership — and for the right fit, real rewards as the team grows.

One hard floor: you have shipped something to production that people other than you depend on, and you have been on the hook at the moment it broke. If that hasn’t happened to you yet, this is the wrong seat — come back when it has.

The workWhat you’d own.

Our goal is deliberate: deliver exceptional results for our clients, exclusively — we are building the AI growth partner of the top companies in this space, and we prioritize exceptional delivery over marketing because that is what compounds.

  • UI first: Next.js · React — the client experience
  • Node
  • Supabase / Postgres
  • Claude API · agent loops · tool calling
  • Voice: ElevenLabs · Telnyx · Twilio
  • n8n
  • Vercel

Plus a verification layer before anything reaches a client — deterministic code that checks the model’s work. Including the boring parts: CRM APIs, webhooks, telephony quirks. Every real system has glue — ours too.

  • The client request queue

    across active accounts — triage, fix, ship, verify, against a response SLA. The lane we most need covered, and the one we most want to shrink.

  • Full-stack feature work on live client apps

    Next.js and React on the front, Node and Supabase behind it. Real users, real data, shipped weekly.

  • The LLM surface

    agent loops with tool calling, structured output that has to be reliable, and model swaps measured against quality rather than guessed at. One platform costs ~$3k/month in model spend; finding the saving without losing accuracy is a real project waiting for you.

  • The verification layer

    the code that checks the model’s output before a client ever sees it. Deterministic, boring, and the most important thing we ship.

  • The interface

    you’ll get a hand-drawn wireframe and a few references, not a Figma file. Mid-level UI craft is enough; the bar is that a client can use what you build without a training call.

  • Deploys, and the boxes they run on

    plus automating your own lane: triage flows, runbooks, self-serve fixes. We want the queue to shrink, not to be lived in.

How we work

Full-time, focused. Independent contractor, no client work elsewhere — personal projects and open source stay yours. Async by default: short Looms, clear written updates, one weekly sync.

Tooling provided. Claude Code and the AI stack we use daily — you get what we use.

Nothing is “done” until it's verified. Our worst incidents were silent successes: an import that reported success while a third of the values were quietly zero; a pipeline that reported healthy for a week after it stopped. You count what you expect.

You work directly with Luke, our Head of AI Solutions — and have a direct line to the founder.

Compensation & working hours$2,500–3,000 / month

Based on your track record. Full-time contractor, paid monthly in USD via international transfer, reviewed at 90 days — it grows with the responsibility you absorb.

How working time actually works here. We set goals together each week, and the goals are what we track — not hours. The daily overlap window exists for one reason: so we can communicate and build together in real time.

~40 hours a week is the honest benchmark for the current workload — a reference, not a clock to punch. Hit the week's goals in less: good for you. If a goal genuinely needs more, we expect you to put in more. It runs on the goals we set together and the speed you work at.

We are not looking for clock-watchers. We're looking for entrepreneurial spirit — people who care that things progress and that clients get exceptional results, because that is what the upside here is tied to.

9:00–17:00 CET14:00–22:00 ICT03:00–11:00 ET

These aren't your working hours — they're the anchor (±1h with daylight saving). At least 4 hours of your day land inside that window, every workday. The rest you organize yourself. What we measure is the queue: response SLA and what ships.

Fit checkWho this is for — and who it isn't.

A fit

  • AI-native, with a floor under you.Ships with AI agents as the default workflow — and reads the logs, the database, the docs when the agent stalls.
  • Verification-disciplined.Distrusts “it ran.” Opens the page, looks, and counts what they expect.
  • Works from a sketch.A wireframe and a couple of references — and comes back with something in the browser.
  • A glue engineer.One request touches a React page, a Postgres function, a model call, and a webhook — and connects them all.
  • Biased to simple.The version that works ships before the version that impresses.
  • Entrepreneurial, not clock-watching.Cares that things progress and clients get exceptional results — the clock isn't the point.
  • Careful with production and client data.Access is trust. Paranoid in the good way.
  • Applies with specifics.Real links, real numbers, things they actually lived through.

Not a fit

  • AI-dependent, no floor.Debugging ends when the AI gives up — if the agent can't fix it, neither can they.
  • “It ran” means “it's done.”Marks work complete without ever opening the page to look.
  • Needs a full spec to move.Waits for a Figma file and a ticket that spells out every step.
  • A single-layer specialist.Comfortable inside one framework — hands everything around it back.
  • Biased to impressive.Overbuilds — the clever architecture matters more than the working product.
  • Watches the clock, not the outcome.Counts hours, not results — done at six whether it shipped or not.
  • Casual with production.Tests in prod, shares credentials, shrugs at client data.
  • Sends generic applications.A mass-produced, model-written pitch with no specifics — we can tell.

Apply

We’re not actively hiring right now — we’re mid-interview on the current round. This goes to our talent pool for future roles: no timed exercise, nothing to do from here. If there’s a fit, we reach out.

A few real questions. Answer them like an engineer.

Every application read. Reply within 5 days, either way. Then an interview, then a paid trial on real work. Decision within 48 hours.

Join the talent pool →

Frequently Asked Questions

What does success look like at 90 days?

The queue runs on SLA without anyone chasing you, and your first end-to-end build is in production. Ninety days is also when the first compensation review happens — it's a real checkpoint, not a formality.

What tools do I get?

Claude Code and the AI stack we use daily — provided, on us, from day one. You bring your own machine and a connection you can take calls on; every subscription the work needs is ours to pay, not yours.

How do I get paid?

Monthly, in USD, by international transfer (Wise or similar), against your invoice. The paid trial is paid out within a week of finishing it, whatever the hiring decision is.

How does the paid trial work?

A real build on Client OS, our internal operations platform — typically a section of it, scoped from half a day up to a few days, at an agreed fixed amount. Real codebase, real review, zero client risk. You see how we actually work; we see how you actually ship. Decision within 48 hours of wrapping up.

How much time off can I take?

Plan around 15–20 working days a year, plus your local public holidays. No approval ritual — but there are rules: flag multi-day time off at least two weeks ahead, hand off anything open in the queue before you go, and never go dark on an active client SLA. Time off doesn't touch your monthly rate as long as the outcomes are covered; holidays go on the shared calendar.

What availability is actually expected?

At least 4 hours of your day inside 9:00–17:00 CET, reachable during that window on normal workdays, and on top of the queue you own. Outside the window, nobody pings you expecting an instant answer. We don't track hours — we notice outcomes, response times, and whether things ship.

What does full-time actually mean here?

This is your only professional commitment — no other clients, no agency side-work; personal projects and open source stay yours. In practice that's a standard ~40-hour week. We don't track hours, we track what ships — but the queue doesn't run on twenty.

What if it doesn't work out?

Thirty days' written notice, either direction, no drama. You invoice through the notice period; we ask for a clean handoff of anything you own. And the 90-day review is where honest conversations happen first — nobody gets surprised.

How many meetings?

One weekly sync, plus the daily overlap window for pairing and handoffs. Async by default: short Looms and written updates instead of status calls. Client-facing calls only when you're genuinely needed — rare, and never sprung on you.

Why is this seat open?

First hire. The founder and the Head of AI Solutions can't be the whole delivery team anymore — that's the point.