Senior Researcher – Agentic Safety & Alignment

You will be joining a world-leading technology company at their R&D site in Helsinki.

We are seeking a high-level scientific expert to lead initiatives in securing and ethically aligning autonomous AI systems. This position focuses on moving beyond standard model safety to address the complex challenges inherent in agents that perform multi-step reasoning, plan tasks, and interact with external tools.
 

Responsibilities

• Advancing AI Alignment: Refine and deploy sophisticated methodologies—such as Constitutional AI, RLHF, and RLAIF—specifically optimized for autonomous workflows and complex chain-of-thought logic.
• Proactive Vulnerability Assessment: Engineer automated red-teaming frameworks and specialized “safety sandboxes” designed to identify and mitigate failure modes unique to agentic systems
• System Hardening: Build robust defensive layers to shield models from adversarial exploits, including prompt injections and jailbreaking attempts that target planning and tool-use functions
• Transparency and Auditing: Develop tools to decode the “black box” of agentic logic, ensuring that the reasoning behind every decision is auditable and transparent.
• Theory-to-Code Pipeline: Translate the latest theoretical breakthroughs from AI safety literature into scalable, high-performance software frameworks ready for production.

Qualifications

• PhD in Computer Science, Machine Learning, Deep Learning, Mathematics, or a relevant  field.
• Proven research track record in generative AI safety and alignment, with deep technical knowledge of LLMs, neural networks, and reinforcement learning.
• Advanced skills in Python, C++, or Java, coupled with extensive experience using major ML libraries like PyTorch or TensorFlow.
• Practical familiarity with agent-centric frameworks, such as LangChain, OpenClaw, or AutoGPT.
• Highly valued background in neural network interpretability, value alignment, model robustness, and editing techniques.
• Prior contributions to AI ethics, safety, or trust-and-safety initiatives are strongly preferred.
• Consistent history of publishing impactful work at premier AI conferences (e.g., NeurIPS, ICML, ICLR, CVPR, or AAAI).
• Success in prestigious international competitions, such as the ICPC, IMO, or IOAI, is highly regarded.
• Drive to work effectively within multicultural teams and a passion for challenging traditional paradigms in the AI field.

Benefits

• Annual Bonus
• Relocation Package
• Lunch Vouchers
• Equipment (laptop, etc.)