onsite
Research Engineer, Frontier Safety Mitigations, DeepMind
Research Engineer
Research Engineer focused on developing frontier safety mitigations for AI systems, applying adversarial robustness, anomaly detection, and reinforcement learning techniques using Python and deep learning frameworks.
About the role
Key Responsibilities
- Design and implement novel algorithms to detect and mitigate adversarial attacks on large‑scale AI models.
- Develop robust anomaly‑detection pipelines that monitor model behavior in real‑time and trigger safety interventions.
- Conduct rigorous empirical studies to evaluate the effectiveness of safety mechanisms across diverse environments.
- Collaborate with cross‑functional research teams to integrate safety mitigations into existing AI pipelines.
- Publish findings in top‑tier conferences and contribute to open‑source safety tools.
Requirements
- Ph.D. or strong research experience in machine learning, with a focus on adversarial robustness or AI safety.
- Proficiency in Python and deep learning libraries such as TensorFlow or PyTorch.
- Hands‑on experience building and evaluating adversarial attacks and defenses.
- Solid background in statistical anomaly detection and reinforcement learning methods.
- Track record of publishing high‑impact research papers or open‑source contributions.
Skills
pythontensorflowpytorchreinforcement learning