AI Engineer (m/f/d) for Spec-Driven Development with Coding Agents
iseremo GmbH
/job-details/13145
**AI Engineer (m/f/d) for Spec-Driven Development with Coding Agents**
Coding Agents · Kotlin/Quarkus · Angular · Tirana, 100% remote · Full-time
**What this is**
We build software with coding agents. A spec gets written and reviewed, Claude Code implements against it, and AI reviews every PR in CI. We're looking for someone who has worked this way, makes the loop reliable, and ships the features behind it. Fully remote, you work from wherever you want in Albania or nearby.
**The three things that matter, in this order**
1. **Coding with agents.**
You've built software spec-first with tools like Claude Code, not just autocompleted with them. You know how to write specs an agent can execute, how to structure repository context files, rules and patterns, and how to wire MCP integrations (Jira, Confluence, browser automation) and AI PR review into the workflow.
2. **Kotlin + Quarkus.**
Kotlin 2.3, Quarkus 3 with CDI, JDK 25. MongoDB, Kafka with Avro, REST with OpenAPI. Maven multi-module, GitHub PRs with a merge queue.
3. **Angular.**
Angular 22, TypeScript, RxJS, NgRx in an Nx monorepo. Jest and Playwright.
**What you'll do**
- Own the context layer that keeps our agents on track: specs, rules, repo instructions.
- Ship features end to end in Kotlin/Quarkus and Angular, spec first.
- Build our first LLM-powered product feature.
- Put evals, guardrails and data-privacy handling in place before anything touches customer data.
**You should bring**
- Hands-on experience building software with coding agents: Claude Code or comparable harnesses, custom agent setups, subagents, hooks, MCP servers. You know what makes an agent drift and how to stop it.
- You keep up with the space. New harness features, model releases, context-engineering patterns: you try them in the first week, not the first year.
- Spec-writing skills. You can turn a vague ticket into a spec that both a reviewer and an agent can execute without guessing.
- Production experience with LLM features: prompting, structured output, tool calling, RAG, evals.
- Solid testing habits (JUnit 5, MockK, Playwright or similar).
- Enough AWS, GitHub Actions and Docker to not be lost. You don’t run the platform.
- Working English. Our team is spread across Germany and the Western Balkans.
**Nice to have**
OpenTelemetry, Keycloak/OIDC, Camunda DMN, ArchUnit, architecture decision records.
Coding Agents · Kotlin/Quarkus · Angular · Tirana, 100% remote · Full-time
**What this is**
We build software with coding agents. A spec gets written and reviewed, Claude Code implements against it, and AI reviews every PR in CI. We're looking for someone who has worked this way, makes the loop reliable, and ships the features behind it. Fully remote, you work from wherever you want in Albania or nearby.
**The three things that matter, in this order**
1. **Coding with agents.**
You've built software spec-first with tools like Claude Code, not just autocompleted with them. You know how to write specs an agent can execute, how to structure repository context files, rules and patterns, and how to wire MCP integrations (Jira, Confluence, browser automation) and AI PR review into the workflow.
2. **Kotlin + Quarkus.**
Kotlin 2.3, Quarkus 3 with CDI, JDK 25. MongoDB, Kafka with Avro, REST with OpenAPI. Maven multi-module, GitHub PRs with a merge queue.
3. **Angular.**
Angular 22, TypeScript, RxJS, NgRx in an Nx monorepo. Jest and Playwright.
**What you'll do**
- Own the context layer that keeps our agents on track: specs, rules, repo instructions.
- Ship features end to end in Kotlin/Quarkus and Angular, spec first.
- Build our first LLM-powered product feature.
- Put evals, guardrails and data-privacy handling in place before anything touches customer data.
**You should bring**
- Hands-on experience building software with coding agents: Claude Code or comparable harnesses, custom agent setups, subagents, hooks, MCP servers. You know what makes an agent drift and how to stop it.
- You keep up with the space. New harness features, model releases, context-engineering patterns: you try them in the first week, not the first year.
- Spec-writing skills. You can turn a vague ticket into a spec that both a reviewer and an agent can execute without guessing.
- Production experience with LLM features: prompting, structured output, tool calling, RAG, evals.
- Solid testing habits (JUnit 5, MockK, Playwright or similar).
- Enough AWS, GitHub Actions and Docker to not be lost. You don’t run the platform.
- Working English. Our team is spread across Germany and the Western Balkans.
**Nice to have**
OpenTelemetry, Keycloak/OIDC, Camunda DMN, ArchUnit, architecture decision records.
