Careers
We hire slowly, and on purpose.
Derkins Labs is a small, senior team, and we add to it one person at a time. Four seats are open. Everything about the job that we can put in writing is on this page, including what it pays.
Roles we hire for
Four seats, all open.
- Engineering · Remote (US / EU) or New YorkOpen
Senior AI Engineer, Integration
Base $185,000 to $235,000 · plus profit share
Own an engagement end to end: scope the opportunity with the client, build the LLM features and agents, wire them into their CRM, ERP, or data pipeline, and ship to production with evaluation in place.
What we look for
- Six-plus years shipping production software, with real LLM work in the last two
- Comfortable in a client's codebase and data model on day one
- Writes clearly: scoping docs, runbooks, and handoff notes are part of the job
- Engineering · Remote (US / EU), with client travelOpen
Forward-Deployed Engineer
Base $170,000 to $220,000 · plus profit share
Embed with a client team for the pilot phase. Sit with the people who do the work, turn what you learn into working software fast, and leave them able to run it.
What we look for
- Full-stack fluency: TypeScript, Python, SQL, and whatever the client runs
- Patience for messy real-world data and legacy systems
- Judgment about when not to build something
- Engineering · Remote (US / EU)Open
Applied Evaluation Engineer
Base $165,000 to $210,000 · plus profit share
Make quality measurable. Design evaluation harnesses, guardrails, and monitoring so clients know how their AI systems behave before and after rollout.
What we look for
- Hands-on experience evaluating LLM outputs at scale, not just reading about it
- Statistical literacy and a healthy suspicion of demos
- Experience with retrieval, agents, or document pipelines
- Client work · New York or Remote (US)Open
Engagement Lead
Base $150,000 to $200,000 · plus profit share
Run the audit and the client relationship. Map workflows, rank opportunities by ROI, and keep engagements honest about scope, cost, and risk.
What we look for
- Operations or consulting background with a technical spine
- Has run projects inside logistics, finance, healthcare ops, or manufacturing
- Can tell a CFO what the work will cost and what it will return
The work
A week in the audit phase, more or less.
- Monday
Audit interviews at a third-party logistics operator. Four hours watching dispatchers requote loads by hand from emailed PDFs. You write down every time someone opens a second window.
- Tuesday
Map their transport management system and the spreadsheet it secretly runs on. Find the field that three departments use for three different things.
- Wednesday
First pass at a quote-extraction pipeline against two hundred real emails. Half of it is dealing with attachments named 'scan0001.pdf'.
- Thursday
Build the evaluation set with the ops lead. Forty cases where two experienced humans disagree about the right answer. Those are the cases that matter.
- Friday
Write the two-page plan the CFO will read on Sunday night. One number per opportunity, one line that says what not to build.
The pilot and integration phases look more like engineering and less like anthropology. The ratio flips, the client stays the same, and you are still the person they call.
The deal
What we pay, and what comes with it.
New York pay, wherever you live
Salary bands are benchmarked to New York and published on every role. Where you sit does not change the number.
Profit share after a year
Every engagement contributes to a pool that is split among the people who did the work. It is written into the offer, not promised at the holiday party.
Twenty-five days off, plus yours
Twenty-five days of leave on top of the public holidays where you live. We track that you take them.
No on-call
We hand production to the client with monitoring and runbooks. Their systems, their rotation. You sleep.
Your machine, your tools
The laptop you want, replaced every three years. A learning budget you spend without asking, up to four thousand dollars a year.
Four hours of overlap
Core hours are four a day in a window that spans US East and Central Europe. The rest of the day is yours to arrange.
The interview
Five questions we will ask you.
- 01
Tell us about a pilot that did not ship. What killed it?
We want to hear you name the real cause, not the diplomatic one.
- 02
Walk us through an evaluation set you built. How did you pick the cases?
'We tested it a lot' is not an answer. Forty hard cases with provenance is.
- 03
A client wants an agent that approves refunds on its own. What do you say?
There is a right answer, and it starts with a question back to them.
- 04
What is a system you would refuse to build?
Everyone has a line. We want to know where yours is before a client finds it.
- 05
Show us the ugliest code you are proud of.
Production software has scars. We would rather see them than a tidy toy project.
You should have questions for us too. Ask about the last engagement we turned down, how the profit share is calculated, and what happened the last time a pilot failed. We will answer all three.
How hiring works
Four steps, two weeks, no puzzles.
- 01
Intro call
Thirty minutes. We tell you how the firm works; you tell us what you want to build. Either side can stop here.
- 02
Work sample
A paid, time-boxed problem drawn from real engagement work. No whiteboard puzzles.
- 03
Working session
Two hours with the team walking through your sample and a live scoping exercise.
- 04
References and offer
We call two people you have shipped with, then make a written offer. Two weeks end to end.
No cover letter. One thing you shipped.
Tell us the role, one thing you have shipped and what it did for the business, and where to find your work. We reply to every application within five business days.