AI Engineer at the Defense's AI Center

Can you explain attention on a napkin, train a transformer from scratch, and build the evaluations that reveal whether it actually works? And will you use your skills on something that really matters? Then we have a mission, access to Denmark's national AI supercomputer, and a place for you, whether you have just submitted your thesis or have a few years under your belt.

About us

AI is already redefining the boundaries of society, security policy, and military tasking. The future battlefield will increasingly be characterized by data, autonomy, and faster decision-making processes. The Danish Defence AI Center (FAIC) has been established as a new unit under the Defence Command to ensure that AI is translated into real Danish and allied defence capability.

FAIC sets the direction for the Defence's use of AI – from doctrine to operational reality. We build institutional competencies, enter partnerships with NATO, allies, academia, and industry, and define the operational needs and principles from which Defence AI solutions are developed and procured – in close collaboration with the Danish Defence Acquisition and Materiel Organization (FMI), so procurements and deliveries are grounded in soldiers' needs and the Defence's core tasks.

Our collaboration with Gefion, Denmark's national AI supercomputer, demonstrates the working culture: open environments enable learning and pace, while closed frameworks ensure military applicability where security, classification, and operational relevance are critical.

FAIC is the place where the Defence's AI future is shaped. We are seeking experienced profiles who want to contribute from day one.

About the role

As an AI Engineer in the Defence AI Center you will work on developing, adapting, and evaluating AI models to meet the Defence's operational needs. You will train and fine-tune models on Gefion and work with everything from transformer architectures, attention mechanisms and optimization to evaluating models' precision, robustness, and security. You will develop the tools, evaluation methods, and safety mechanisms that make it possible to assess whether a model is ready for operational use.

At the same time, you will build data pipelines, test tools, and software integrations and collaborate closely with developers, platform engineers, and operational users to translate research into solutions that work in practice.

You will become part of a cross-disciplinary environment with a steep learning curve, where both new graduates and experienced professionals can develop professionally and contribute their skills.

The role is for you if you are curious about the latest AI research, motivated to understand why models behave the way they do, and want to use your technical knowledge to strengthen the Defence's mission.

Your key tasks

Model development and training

You will train, fine-tune, and adapt AI models for Defence tasks and data, including language models, multimodal models, and models for sensor and image data. You will experiment with transformer architectures, attention mechanisms, and other effective model structures and apply methods such as supervised fine-tuning, preference optimization, and reinforcement learning to shape model behavior. On Gefion and the AI Center's other GPU resources you will optimize training runs with a focus on efficiency and stability.

You will also build and quality-check training and evaluation data, including synthetic data, focusing on quality, traceability, and classification.

Evaluations, guardrails, and trust

You will develop evaluation methods and benchmarks that measure models' precision, robustness, and reliability in relevant operational scenarios. You will implement guardrails that support safe and predictable model behavior, also against prompt injections and other adversarial attacks. Through red teaming and failure analysis you identify weaknesses and document results so there is a solid basis for assessing whether models are ready for use. You will also contribute to explainability and responsible use of AI in accordance with Defence and NATO principles.

Inference and deployment

You will develop inference harnesses and evaluation pipelines that enable testing, comparing, and validating models systematically and reproducibly. You will optimize models for practical use through quantization, distillation, and efficient serving, so they can run on limited hardware and in isolated environments. In collaboration with developers and platform engineers you contribute to bringing models safely from development to classified operating environments following the principle develop low, deploy high .

Interdisciplinarity and breadth

You will develop components, services, and tools that support model development, including data loading, APIs, integrations, and user interfaces for testing and demonstration. You will maintain data pipelines and test data that make the development process more efficient and reliable, and you will collaborate with developers, platform engineers, data specialists, and operational users to bring solutions from idea to use. You take responsibility for the tasks that arise along the way, continuously acquire new skills, and share your knowledge with the team.

Research and collaboration

You follow international AI research and translate relevant results into concrete solutions for the Defence. Based on mathematical modeling, statistics, and optimization you explore complex problems and assess which models and methods best solve the task. You collaborate with research environments, industry, NATO, and allies on model development and knowledge sharing. Additionally, you prepare technical analyses and recommendations that support Defence decisions, and contribute to building the organization's AI competencies through knowledge sharing, professional mentoring, and coaching.

We offer a job with a mission: You help the Defence adopt AI, gain access to Denmark's national AI supercomputer, work on problems you won't find elsewhere, and become part of a team where graduates and experienced professionals learn from each other and build the Defence's own AI competence from the ground up.

About you

You hold a master's or PhD in machine learning, computer science, mathematics, physics, statistics, or another corresponding quantitative field. You are a recent graduate or have experience from research or practice where you have worked on training and fine-tuning deep learning models – preferably language models or other transformer-based models.

You have a solid understanding of transformer architectures, attention mechanisms, and the mathematics behind modern deep learning, including linear algebra, probability, and optimization. You also have experience with evaluating models, inference harnesses, and guardrails and you take a critical view of what benchmarks really measure. You are strong in Python and PyTorch or similar frameworks and write code that is reproducible and usable by others.

You thrive in a cross-disciplinary environment and are eager to work broadly with AI – including tasks around the models such as component development, data processing, and integrations. You are motivated by a clear mission and want to use your technical skills to strengthen Denmark's and NATO's defence. You are proficient in both Danish and English in a professional context.

It is advantageous, but not required, if you also have experience in one or more of the following areas:

• Distributed training: Training on GPU clusters with e.g. Slurm, FSDP, DeepSpeed, or Megatron.
• Model optimization: Preference optimization and reinforcement learning, e.g. DPO or RLHF.
• Efficient inference: Quantization, distillation, or tools such as vLLM, TensorRT, and CUDA.
• AI security: Mechanistic interpretability, red teaming, or other methods to investigate and improve model safety.
• Multimodal models: Models for sensor, image, or geodata.
• Research and open source: Publication of research or contributions to open source projects in machine learning.
• Defence and government: Experience from defence, intelligence, or government contexts, including work in classified environments.

Employment terms

The appointment is governed by the applicable collective agreement between the Danish Ministry of Finance and a number of academic associations for academics in the state (Academic Agreement).

Qualification and functional supplements are negotiated based on relevant experience and competencies. In addition, there is an employer-funded pension scheme attached to the position.

The workplace is in Lyngby

As an employee under the Ministry of Defence, you are subject to certain special terms. You must be able to obtain and maintain a security clearance during the entire employment period.

You have a duty as an employee to perform service outside the country, including at units in international operations with a view to participating in conflict prevention, peacekeeping, peacebuilding, humanitarian and other similar operations.

Contact and application

If you want to hear more about the position, you are welcome to contact Rasmus Kilmose at telephone 2262 8836.

If you have questions about salary and employment terms you are welcome to contact the Personnel Command, HR consultant Madeleine Moltke Kæmsgaard at telephone 4031 1094.

The application deadline is 30 November 2026 and interviews will be held on an ongoing basis. The position is available as soon as possible after agreement.

You apply for the position via the link, and we ask you to attach your application, CV and copies of your diplomas. Please note that attachments must be added in PDF format.

In the Defence Ministry's organization we believe that diversity optimizes task solving. Therefore we encourage all interested regardless of background to apply for the position.

About the Defence AI Center

AI is redefining both the battlefield and the way the Defence creates and uses military capability. AI shortens the path from observation to decision and impact and opens up task solutions that were not previously possible. AI is not something that merely supports the Defence's operations — AI is helping to define them. Therefore, AI has a central place in the Defence, and that is why the Defence AI Center (FAIC) has been established as a new unit under the Defence Command.

FAIC sets the direction for how the Defence uses AI, so investments translate into real Danish and allied defence capability. We prioritize operational needs across services and commands in close dialogue with operators, and we build partnerships with NATO, allies, academia, and industry. Our task is to ensure that AI becomes combat power — not technology for technology's sake.

We work AI-first and innovatively enough to change us as reality changes. That means short cycles, close to the operator, and tempo and learning as operational requirements. We develop in open environments to gain speed and insight, and we take solutions into operation in classified environments where security and operational relevance are critical. We have access to some of the Nordic region's best computing, and our infrastructure is built so that AI is part of the solution from the start.

The Defence AI Center is the place where the Defence's AI future is shaped.