About Aithos
The values expressed by AI, and who controls them, will shape how individual people live, work and make decisions.
Aithos is an independent AI research foundation. In our first year, our two-person research team published two academic papers, a preprint, eight research blogs, our results were covered in 70+ news outlets and we generated lasting relationships with governance bodies, such as the Autoriteit Persoonsgegevens. Our mission is to protect human autonomy and pluralistic societies by making AI system values transparent and steerable at the individual level. We also investigate the coordination problems that emerge when increasingly powerful AI systems and agents with different values interact.
We produce research papers for scientific discussion, accessible materials for public debate, and tools to evaluate and steer AI value systems. We aim to turn every piece of research into a benchmark, open (evaluation) tooling or a reproducible methodology. Our outputs are public, reproducible measures that labs, regulators and other researchers use to carry out their own ethical research using our methods, or compare models on the questions we care about.
About the Role
We're looking for a Research Engineer to join our team. Our current research mainly focuses on LLM behaviors in agentic settings over repeated runs, either multi-turn or with pregenerated histories. Our research team focuses on research questions, measurement targets, construct validity, and experimental design. You lead execution, building the evaluations to be reliable, efficient, reproducible, true to the experiment design, and accessible for the public.
You'll be working together closely with our technical lead and research team. On your first day you inherit working research code that implements existing designs and produces results. By the end of your first year, you'll be creating new experiment setups in collaboration with researchers, run new models through our benchmarks within a week of release, be able to answer questions about system architecture, facilitate others to run new experiments in the environment you built, and co-author research papers."
You are responsible for:
-
The software stack. The code, architecture, harnesses, scenarios, scoring and model integrations behind our evaluations. Clean, tested and documented, not one-off research scripts.
-
The runs. Getting new models through our evaluations quickly, cheaply and reproducibly. That means managing API and compute budgets, versioning every run, and making sure the numbers in a paper, a press release or a regulator briefing trace back to a specific commit and dataset.
-
The research tooling. Our moral competence and multi-agent research need experiment infrastructure: agent environments, logging, analysis pipelines. You make it possible for a two-person research team to run studies that would normally need a lab.
You will also be responsible for how our technical credibility shows up in public. Regulators, partner organisations and other evaluation teams will read our code. Open, well-documented tooling that others adopt is one of the ways a small European foundation earns a seat at the table, and it takes judgement to know what to release, when, and with what caveats.
You will also:
- Build tools to enable the steering of AI systems’ values
- Collaborate with our interdisciplinary team to integrate insights from philosophy, social sciences, and computer science
- Help position our work within the broader AI safety and alignment community, identifying where our approach diverges from or complements mainstream thinking
Requirements
What We're Looking For
We don't need decades of experience or a formal background, and we welcome self-taught candidates. We need someone motivated to do impactful work who's comfortable with ambiguity and technical inquiry formed by AI safety concepts, security, philosophy and ethics. You might be:
- A software or ML engineer who has shipped code other people depend on and wants to move into AI safety
- A researcher (PhD, postdoc or independent) whose strongest skill is building the code behind experiments
- An experienced engineer from industry or an AI lab who wants their work to have a more direct impact on AI safety
Required
- Strong software engineering skills in Python: clean, tested, documented code that others can build on, and the ability to implement experiments.
- Hands-on experience with LLMs beyond single API calls: evaluations, agents and scaffolding, prompting or fine-tuning.
- A scientific mindset: you look for alternative explanations and do not over-read a result.
- You treat security as part of the engineering: safe sandboxes for agents that write and run code, and careful handling of data and credentials.
- You use AI coding tools to move fast, and you are disciplined about checking their output.
- Genuine motivation toward AI safety or doing good through impactful work
Bonus
- Experience with Inspect or other evaluation frameworks
- ML publications or preprints (helpful, not required)
- Experience building multi-agent systems or agent simulations
- Open-source contributions, especially evaluation or safety tooling
- Availability to work on-site in Amsterdam
Why Join Aithos?
Most alignment research focuses on technical safety but leaves critical questions out of scope: whose values should AI systems follow? What happens when these conflict? How should alignment work when AI systems interact in complex ecosystems? Who has the authority to decide? At Aithos, we start from the fact that people disagree about values, and build alignment work that holds up under that disagreement instead of assuming it away. You'll work on genuinely novel problems with a team that values intellectual honesty and diverse perspectives, without the constraints of traditional academia or industry.
We're based in Amsterdam and work together in person at least two days a week. As a small foundation started in 2025 and aiming for significant growth in coming years, we offer autonomy, the chance to shape research directions, and the opportunity to contribute to work that challenges how AI systems represent human values while working alongside a team of like-minded researchers. Salary: €65,000 to €85,000 gross per year depending on experience, plus 8% holiday allowance.
Application process
To apply, email your resume and a brief motivation letter to hiring@aithos.org by 31 October.
We will review applications in the first week of November. Shortlisted candidates will be invited to take part in three steps:
- A 30-minute screening call. This will take place online.
- A take-home assignment based on a real Aithos evaluation task, taking approximately three hours.
- A 60-minute technical conversation with our Technical Lead. At this stage, we prefer to meet candidates in person at our Amsterdam office.
After the final conversation, we will follow up with our decision and next steps.