Principal Research Scientist, Trust & Safety Alignment

Principal Research Scientist – Trust & Safety Alignment

About the Role

We are seeking a Principal Research Scientist to lead advanced research in AI alignment, trust and safety, and foundation model reliability.

The role focuses on improving the safety, robustness, controllability, and interpretability of large language models, vision-language models, and agentic AI systems. Research areas include continuous and post-training, reinforcement learning, preference optimisation, model editing, machine unlearning, interpretability, controlled learning, and agentic AI safety.

This position is suited to an experienced research leader who combines deep technical expertise with a strong track record of advancing the state of the art and translating research into scalable, production-ready systems.

Key Responsibilities

• Lead research and development of advanced alignment and safety methods for large language models, vision-language models, and agentic AI systems.

• Develop techniques for continuous training, post-training, reinforcement learning, preference optimisation, and controlled model adaptation.

• Research methods for aligning model behaviour with safety requirements, human preferences, and intended system objectives.

• Investigate new neural network architectures, learning algorithms, and optimisation techniques for foundation models.

• Develop approaches for neural network editing, model behaviour modification, machine unlearning, and controlled knowledge updates.

• Advance research in interpretability, explainable AI, mechanistic analysis, and reverse engineering of neural networks.

• Develop techniques to understand, diagnose, and modify internal model behaviour.

• Define long-term research directions and contribute to strategic roadmaps in AI safety, alignment, robustness, and trustworthy AI.

• Identify emerging research opportunities and establish new technical programmes.

• Design and run hands-on experiments to validate novel research ideas.

• Translate successful research outcomes into robust, efficient, and scalable systems.

• Lead collaborations with universities, research institutions, engineering teams, and external research partners.

• Provide technical leadership and mentorship to researchers and engineers.

• Communicate complex research findings clearly to both technical and non-technical stakeholders.

Essential Requirements

• PhD in Computer Science, Artificial Intelligence, Machine Learning, Deep Learning, Mathematics, or a related technical field.

• Eight or more years of relevant research experience in artificial intelligence, machine learning, AI safety, security, or a closely related area.

• Strong research track record with evidence of significant technical contributions.

• Deep expertise in large language model or vision-language model continuous training, post-training, and alignment.

• Strong knowledge of reinforcement learning, preference optimisation, RLHF, adversarial training, or related techniques.

• Proven experience with neural network architecture, algorithm design, model optimisation, and foundation model development.

• Experience with large-scale model training, adaptation, evaluation, or deployment.

• Strong understanding of neural network editing, model interpretability, explainable AI, or model reverse engineering.

• Ability to independently define and lead complex research programmes.

• Strong experimental design, analytical, and problem-solving skills.

• Ability to translate research concepts into practical and scalable implementations.

• Strong written and verbal communication skills.

Preferred Qualifications

• Experience with neurosymbolic AI, structured reasoning, knowledge grounding, or hybrid neural-symbolic systems.

• Experience using structured knowledge to improve factuality, model robustness, or reasoning quality.

• Research experience in AI safety, reasoning, robustness, security, fairness, transparency, or abuse and risk detection.

• Experience with machine unlearning, model editing, continual learning, or controlled adaptation.

• Experience with agentic AI systems, tool-using models, planning systems, or multi-step reasoning.

• Experience developing defences against adversarial attacks, model exploitation, prompt injection, or other AI security threats.

• Experience with mechanistic interpretability, activation analysis, probing, representation analysis, or internal model diagnostics.

• Experience building or evaluating scalable safety and alignment systems in production environments.

• Strong collaborations with academic or industrial research institutions.

Research Track Record

A strong publication record at leading AI, machine learning, computer vision, or natural language processing conferences is highly valued.

Relevant venues may include:

• NeurIPS

• ICML

• ICLR

• AAAI

• ACL

• CVPR

• ICCV

• EMNLP

• NAACL

Leadership Profile

The successful candidate will combine deep technical research expertise with the ability to establish and lead ambitious research programmes.

You should be comfortable defining research direction, identifying high-value technical problems, mentoring researchers, and maintaining hands-on involvement in experimentation and system design.

The role requires the ability to bridge fundamental research and practical deployment, ensuring that new techniques can progress from early-stage research into robust, scalable AI systems.

Position Details

Location: Helsinki, Finland
Employment type: Permanent, full-time

Why Join

• Help define long-term research direction across AI alignment, trust and safety, machine unlearning, and advanced foundation models.

• Work on research that spans both fundamental AI problems and real-world deployment.

• Lead work across reinforcement learning, model alignment, interpretability, model editing, and AI safety.

• Collaborate with experienced international researchers, engineers, universities, and research organisations.

• Lead multi-year research initiatives and external collaborations.

• Contribute to the development of safer, more robust, and more trustworthy AI systems operating at scale.