TradeLink builds agentic platforms that lift European supply chains to the next level. Intelligent AI agents take over the routine, while the important decisions and priorities stay with your team. A network connecting companies, suppliers and carriers ensures maximum transparency, reduces lost time, and creates a secure, trusted process. With TradeLink, you're always in command of your logistics — not chasing it.
Quick overview
SeniorityMid Senior
Employment typeFull Time
Work modeHybrid
LocationMunich, Germany
Posted7 hours ago
RubyGitGitLab CIRailsGenerative AIReactPostgreSQLTypeScriptCI/CD
Job description
About TradeLink
Code is becoming cheap. Attention, judgment and trust are not
Coding agents can produce an enormous amount of software.
If an agent can open 20 merge requests but senior engineers still need hours to understand and verify every one, we have not removed the bottleneck. We have moved it.
TradeLink wants to redesign the software-development system around this reality.
We are looking for a Staff Software Engineer — Agentic Development to build the harness that lets a small engineering team safely ship at agent speed.
What you'll do
Your job is not to introduce another AI coding tool. Our engineers already use them.
Your job is to answer the harder question: What should an engineering organization look like when implementation is increasingly done by agents, but humans remain responsible for architecture, judgment and quality?
You will own:
-
The coding-agent harness: context, tools, instructions, environments and feedback loops.
-
Making our repository easier for agents and humans to understand and change.
-
Automated verification: tests, E2E, browser checks, static analysis and other evidence.
-
The MR handoff: scope, summaries, risks and proof that helps a human review quickly.
-
CI and disposable environments agents can use reliably.
-
Risk-based human approval gates.
-
Metrics for agent success, rework, review effort, lead time and escaped defects.
-
Turning great individual AI workflows into team-wide engineering capabilities.
The goal is not more generated code. It is more safe customer value per engineer with less human cognitive load.
What you bring
We are looking for a very strong software engineer who has independently gone deep on coding agents.
-
Roughly 7+ years as a software engineer, or equivalent evidence that you operate at Staff level.
-
You have spent years writing, reviewing and operating production software. You know what good engineering looks like before AI enters the picture.
-
Claude Code, Codex or equivalent coding agents are already central to how you work.
-
You do more than prompt: you have built your own harnesses, scripts, tool/MCP integrations, repository instructions, verification loops or multi-step agent workflows.
-
You think deeply about context engineering, testing, CI, code review and human cognitive load.
-
You can personally review a difficult MR and spot architectural problems, bad abstractions and weak tests.
-
You have experience with developer tooling, CI/CD, testing infrastructure or engineering productivity.
-
You can work inside the product codebase as well as on tooling around it.
-
You experiment constantly and can show us approaches that failed as well as those that worked.
-
You care about increasing the output of the whole engineering organization, not just your own productivity.
- Fluent German and English language skills are required
Our application stack includes Ruby on Rails, React, TypeScript, PostgreSQL and GitLab CI. Exact stack experience is useful; deep production-engineering judgment is essential.
Backgrounds that could fit
You might be:
-
A Staff/Principal Product Engineer who became extremely AI-native.
-
A Developer Productivity or DevEx Engineer with serious application-engineering credentials.
-
A Founding Engineer from an AI-native startup.
-
An engineer building internal coding-agent infrastructure at a high-performing software company.
If your experience with coding agents is mostly autocomplete, generating individual functions or asking ChatGPT questions, this role is probably too early for you.
This probably isn't for you if
-
Your favorite AI productivity metric is generated lines of code.
-
You think another AI reviewer automatically solves review overload.
-
You want agents autonomous before they are verifiable.
-
You love AI tooling but no longer enjoy product code.
-
You cannot confidently review architecture and code without AI telling you whether it is good.
How we think about AI
Before becoming obsessed with coding agents, you became a strong engineer. That matters.
AI-native does not mean accepting everything an agent writes. Almost the opposite.
You know how to delegate aggressively while remaining accountable for the outcome. You think about context quality, feedback loops, tool access, verification, failure recovery and human attention.
The goal is not maximum autonomy.
The goal is maximum trustworthy leverage.
Why TradeLink
We are in the middle of this transition, not writing a strategy deck about it.
Our engineers already use coding agents. Implementation speed is increasing. Human review and cognitive load are becoming visible constraints.
Because TradeLink is a small product organization, a successful improvement can change how the entire engineering team builds software — while staying connected to real product and customer problems.
How we interview
Bring your AI workflow.
We want to understand the most sophisticated engineering task you currently delegate to an agent and what you deliberately keep human. In the working session, AI tools are allowed and expected. We care about delegation, context, verification, risk judgment and whether another engineer could review your final change quickly.
One question before you apply: If a coding agent could produce 10× more changes tomorrow, what becomes the new bottleneck — and what would you build first?