AI Safety Researcher

We are looking for a researcher to further strengthen our work on AI safety. You will work with a cross functional team of highly skilled researchers, engineers, and domain experts on making sure our features are safe and trustworthy. You have a strong technical background and are able to work hands-on with complex systems and data.

What You'll Do
-Working with a cross functional team including Research, Trust & Safety and Engineering.
-Adversarial Testing: Stress test systems, e.g. via red-teaming campaigns, to identify material gaps and produce training data. The work includes for example:
-Working hands on with querying and managing data, automated red teaming frameworks, LLM-as-ajudge, and more
-Benchmarking with similar services
-System alignment: Work with the teams to better align systems with evolving safety policies, focusing on robust and scalable processes. Example work:
-Prompt and context engineering; Preference Tuning; Automatic prompt optimisation
-Producing high quality test and training data
-Preferably work full time during the contract, but part time can be applicable as well

Who You Are
-Essential Safety Experience: Proven experience contributing to safety-related projects or research (e.g., adversarial testing, system alignment).
-Technical Stack: Strong proficiency in Python, Java, and SQL.
-AI Expertise: Hands-on experience with LLMs and prompt/context engineering.
-Academic Requirement: Preferably pursuing or holding an MSc or PhD in an AI/ML-related field, with a focus on safety or agentic systems.
-Plus: Experience working with cross-language models.
-Core Expertise: Safety Research and advanced model alignment techniques.
-Responsibilities: Lead adversarial testing/red-teaming campaigns to identify material gaps, focusing on robust and scalable system alignment (e.g., Preference Tuning, automatic prompt optimisation).

Workplace: Sweden
Start: Feb
Length: 6 months