Principal AI & Agent Systems Engineer (gn) @ Fusion Energy Venture, Munich

This is an Atlantic portfolio venture.

We are building a new path to fusion energy. Every major leap in human civilisation has followed a breakthrough in energy, from fire to steam to electricity, and we believe fusion is the next one. Our bet is that it will not come from building bigger machines. It will come from controlling matter with enough precision that fusion becomes a manufacturable technology, atoms aligned exactly enough that fusion happens under controlled conditions, on a chip. This is a semiconductor-scale path to clean power, and our stance is precision over brute force. We start in physics and simulation, then build the physical proof. 

Reaching a chip-scale approach means working through physics and simulation at a depth that is hard to reach by conventional means, and to do that we need a new class of scientific intelligence. You will work directly with the founders to design that system from the ground up: an AI-driven research platform that can read papers, connect ideas across disciplines, orchestrate simulations, evaluate hypotheses and learn from results. This is not a chatbot or prompt engineering role, and we are not building a foundation model or competing with OpenAI, Anthropic or Google. We believe these systems will need richer representations than today's token-prediction models, so you will help explore new architectures for scientific reasoning, memory, hypothesis generation and autonomous discovery. The better that system works, the faster we get to the physical proof.

• →

  Design multi-agent research systems and the orchestration that ties them together

• →

  Build long-term memory architectures for scientific reasoning

• →

  Create paper ingestion and knowledge extraction pipelines

• →

  Develop scientific reasoning workflows and connect AI agents to simulation environments

• →

  Build autonomous experiment and evaluation loops

• →

  Design retrieval, planning and orchestration systems

• →

  Integrate state-of-the-art LLMs and open-source models

• →

  Develop scalable infrastructure for continuous learning

• →

  Explore next-generation AI architectures for scientific discovery

We care far more about what you have built than about formal credentials. You will thrive here if you enjoy solving problems nobody has solved before, learn quickly and independently, and are comfortable with uncertainty. This role suits someone who wants to help create a new category of scientific intelligence rather than optimise an existing product.

• You have already built real agent systems, with strong experience across several of: multi-agent architectures, tool use and function calling, agent orchestration, planning systems, long-term memory, knowledge graphs, autonomous research workflows, RAG architectures and evaluation frameworks

• You write production-quality software with strong Python skills, and you are comfortable with API design and integration

• You know your way around cloud infrastructure, Docker and containerisation, and databases including vector databases

• You are comfortable with the mathematical ideas this work draws on, such as linear algebra, optimisation, probability, graph theory and dynamical systems

• You do not need to be a theoretical physicist, but you should enjoy working on highly technical scientific problems

• Physics simulations, scientific computing or computational physics

• HPC environments

• Reinforcement learning

• AI for Science

• Quantum computing

• Scientific publishing workflows

• Open-source AI frameworks