About the Role

As Aevum Zenth scales foundational AI models across our 400 global subsidiaries, ensuring robust alignment, interpretability, and ethical deployment is paramount. The AI Safety Researcher will work at the intersection of theoretical computer science, machine learning, and systems engineering to develop next-generation safety mechanisms for large-scale neural architectures.

You will collaborate directly with our Zenth Digital Systems and Advanced Research teams, conducting rigorous evaluation, red-teaming, and alignment research on production AI systems that power everything from autonomous logistics to precision healthcare diagnostics.

Key Responsibilities

• Design and implement novel evaluation frameworks for model capability, alignment, and failure mode analysis.
• Conduct systematic red-teaming and adversarial testing across LLMs, multimodal systems, and agentic workflows.
• Develop interpretable monitoring tools and runtime safety guards for deployment pipelines.
• Collaborate with model architects to integrate safety-by-design principles into training and inference stages.
• Author and present peer-reviewed research on alignment, robustness, and AI governance at top-tier conferences.
• Contribute to internal safety standards, documentation, and cross-divisional AI risk mitigation protocols.

Qualifications

• Ph.D. or M.S. in Computer Science, Machine Learning, Applied Mathematics, Physics, or related field.
• 3+ years of research experience in ML/AI safety, interpretability, or robustness.
• Strong programming skills in Python, PyTorch/JAX, and familiarity with large-scale distributed training.
• Published work in NeurIPS, ICML, ICLR, or safety/alignment workshops.
• Deep understanding of transformer architectures, optimization dynamics, and emergent behaviors.
• Experience with evaluation benchmarks, dataset curation, and automated safety testing pipelines.