Engineering

Agentic Systems Engineer (all genders)

Remote-firstLLM agentsEvals & guardrails

Apply Now Opens the email app All Open PositionsGoes to: Join Us — Talents

[01]The Role

What This Seat Is For.

Design multi-step agent systems for research, outreach, and reporting — with the evals, guardrails, and reliability engineering that make them production-grade. We run agents on real client work; toys don’t survive here.

[02]What You’ll Own

Your Scope From Week One.

Design and build multi-step LLM agent workflows for research, outreach, and reportingOwn evals: measure quality before and after every changeBuild guardrails — input validation, output checks, human-in-the-loop gatesEngineer reliability: logging, retries, fallbacks, cost and latency budgetsProductize proven agents into the overtools suite

[03]What You Bring

The Profile That Wins This Seat.

Tick what’s true. Be honest — we will be.

Hands-on LLM agent experience — Any framework — reasoning matters more than the logo. Strong Python or TypeScript — Production-quality code, not notebooks only. An eval methodology — One you can explain and defend. Production mindset — Observability, failure modes, graceful degradation. GDPR awareness — Careful handling of client and prospect data.

0 / 5Your call — the boxes are waiting. Opens the email app

Preferred, not required

Evaluation tooling in production. Braintrust, LangSmith, or something you built yourself.Retrieval and context engineering. You know why a bigger context window is not an answer.Cost and latency instincts. You have argued about tokens per run and won.Something open source we can read. Code, a write-up, a talk.

[04]The Practical Details

What This Actually Looks Like.

Start As soon as possible

Engagement Part time, Freelance / contractor, Temporary, Paid internship

Where and when Remote-first, core hours overlapping CET

Languages English for the work, German helpful but not required

Travel Rare, a few team days a year in Berlin or Munich

You work with The low-code engineer, the AI revenue engineer, and the mandate teams your agents touch

[05]The First Quarter

Your First 90 Days.

By day 30

Read the running agents end to end, then write the eval that should have caught the last failure.

By day 60

Ship one agent workflow of your own into a live mandate, with guardrails and a cost budget you defend.

By day 90

Own the reliability numbers for the whole fleet: quality, latency, cost per run, and the trend on all three.

Deep work over performative hours, senior review inside live mandates, and scope that grows before the title does. What we offer, and how we hire ➞