Job
- Level
- Senior
- Job Field
- IT, DevOps, Back End
- Employment Type
- Full Time
- Contract Type
- Permanent employment
- Location
- Munich
- Working Model
- Onsite
Job Summary
In this role, you design and operate the AI agent platform, optimize the architecture, integrate tools, and ensure safe autonomy in a productive cloud environment.
Job Technologies
Your role in the team
- Technical owner of Proxima's AI agent platform, turning autonomous agents into reliable everyday colleagues for physicists and engineers.
- Own the platform architecture, technical delivery, and day-to-day operation.
- Scale the platform from an early product into reliable, governed infrastructure used across the company.
- Enable agents to work across knowledge bases, write code, run simulations, and support engineering workflows.
- Work within the Enterprise AI Team Lead's overall vision while owning how that vision is technically implemented.
- Use strong technical judgment to influence and shape the platform roadmap.
- Identify and propose new platform features based on hands-on experience and feedback from internal users.
- Build a platform that engineers and scientists can trust for real, production-level work.
- Own the design, delivery, and operation of our agent platform: the runtime and its primitives: durable sessions, sandboxed execution environments, long-running task execution, memory, and tool integration.
- Turn a rapidly evolving agent ecosystem into stable platform building blocks: evaluate harnesses, models, and protocols as they emerge, and decide what becomes a supported primitive versus a passing experiment.
- Co-design the integration layer that connects agents to internal and external systems, from enterprise tools to our scientific computing stack, and the standards for how tools, skills, and context are exposed to agents.
- Make autonomy safe and governable: permission and approval flows with human-in-the-loop escalation, capability boundaries, audit logging, cost and token controls, and hardening against prompt injection and tool abuse.
- Build the evaluation and observability layer for non-deterministic systems: session tracing, quality evals, and cost/latency visibility, so we know agents work before our users have to find out.
- Operate what you build on Google Cloud with production discipline: reliability, failure handling, and observability.
- Work directly with demanding technical users and feed their feedback and your technical judgment into the platform roadmap you shape together with the team lead.
This text has been machine translated. Show original
Our expectations of you
Qualifications
- You have architected, built, and operated production-grade agentic or LLM systems well beyond the prototype phase.
- You know the current agent ecosystem hands-on: MCP, agent SDKs and CLIs, multi-agent orchestration patterns, sandboxed code execution and have formed opinions on where the abstractions should sit.
- You have run production workloads on Kubernetes and a major cloud (ideally GCP), and know what it takes to keep stateful, long-running services alive.
- You treat evaluation and observability as platform features, not afterthoughts.
- You take security seriously as a design constraint.
- You think in platforms and internal products.
- You are a low-ego, high-clarity communicator who can work directly with expert users and build consensus across teams.
- Exposure to scientific or high-performance computing: batch schedulers, HPC clusters, or simulation-heavy environments.
- Open-source contributions to AI/agent tooling.
- A track record of technical mentorship.
Experience
- You have 5+ years of software engineering experience, with real depth in Python and distributed backend systems.
- You have experience with infrastructure as code (Terraform) and CI/CD.
- Erfahrung mit Model Gateways und Routing.
This text has been machine translated. Show original
What we offer
- This role sits at "L3" of our framework, please inquire during the recruitment process for further information.
This text has been machine translated. Show original
Topics You Will Work On
Job Locations
About Your Employer
Proxima Fusion
Proxima Fusion GmbH, founded in 2023 in Munich, focuses on the development of the world's first commercial fusion power plant. The company employs innovative stellarator technology and modern methods such as machine learning to create market-ready fusion facilities by the early 2030s. As a spin-off of the Max Planck Institute for Plasma Physics, Proxima Fusion benefits from a strong connection to the European research landscape.
Description
- Company Type
- Startup
- Working Model
- Onsite
- Industry
- Power Sector, Economy
Senior Agent Platform Engineer
- Location
- Munich
- Working Model
- Onsite
- Diversity
- Open for all genders
- English Only
- English only required
