AI developer

Levato; Norway; Sur place; NOK 1 000 000 - 1 400 000; Plein temps

Résumé du poste

Levato in Norway is seeking an AI engineer to design and implement the AI agent training pipeline at VANTAGE, combining imitation learning from wargame recordings with large‑scale reinforcement learning inside a physics‑consistent military simulation.

You will develop RL pipelines, implement imitation learning, generate synthetic scenarios with LLMs, build RAG grounding, and ensure explainability with live MCP integration during wargames.

Qualifications

• Strong grounding in reinforcement learning and deep learning.
• Hands-on experience training and evaluating AI models.
• Proficiency in Python and familiarity with modern ML frameworks (PyTorch or equivalent).

Responsabilités

• Design and implement RL pipelines for training tactical AI agents inside the VANTAGE simulation environment.
• Imitation learning from structured human wargame recordings — capturing and encoding operational judgment.
• Synthetic scenario generation using LLMs to create operationally contextualised training environments.
• RAG (Retrieval Augmented Generation) architecture for grounding AI reasoning in doctrine and operational data.
• Explainability and traceability frameworks — ensuring every agent recommendation can be audited.
• Integration with the MCP interface for real‑time agent‑simulation interaction during live wargames.

Connaissances

Reinforcement learning

Deep learning

Python

PyTorch

Description du poste

Train the AI agents that will help NATO commanders make better decisions.

The Role

This is one of the most technically interesting AI engineering roles available in European defence technology. You will design and implement the AI agent training pipeline at the core of VANTAGE — a system that combines imitation learning from human wargame recordings with large-scale reinforcement learning inside a physics-consistent military simulation environment.

The problem you are solving is genuinely novel: how do you train mission‑specific AI agents for military decision support without classified training data? The answer — generating synthetic operational data inside the simulation itself — requires deep AI expertise applied to a domain that matters. Your work will directly influence how allied commanders explore courses of action in complex, contested environments.

What You Will Work On

• Reinforcement learning pipelines for training tactical AI agents inside the VANTAGE simulation environment
• Imitation learning from structured human wargame recordings — capturing and encoding operational judgment
• Synthetic scenario generation using LLMs to create operationally contextualised training environments
• RAG (Retrieval Augmented Generation) architecture for grounding AI reasoning in doctrine and operational data
• Explainability and traceability frameworks — ensuring every agent recommendation can be audited
• Integration with the MCP interface for real‑time agent‑simulation interaction during live wargames

Requirements

Essential: Strong grounding in reinforcement learning and deep learning. Hands‑on experience training and evaluating AI models. Proficiency in Python and familiarity with modern ML frameworks (PyTorch or equivalent).

A strong plus: Experience with imitation learning, multi‑agent systems, or simulation‑based training environments. Familiarity with LLM integration, RAG architectures, or prompt engineering. Interest in explainable and responsible AI.

Who you are: You are driven by hard problems at the frontier of applied AI. You care about building systems that are trustworthy and explainable, not just performant. You want your research to ship into a product used by real military organisations — not remain in a notebook.

About the Company

Levato is a Norwegian defence‑technology company building VANTAGE — the NATO Alliance's most advanced AI‑native digital wargaming and decision‑support platform. We are used by armed forces across Norway, Ukraine, the Netherlands, the United States, and beyond. We are growing fast and looking for exceptional engineers who want their work to matter.