AI Researcher

Description Join Onyx Security and help build the control plane for autonomous AI in a fast-moving, high-impact environment. Onyx is building the control layer that enables enterprises to safely adopt and scale AI agents across the organization, with the visibility, governance, and operational control needed to understand what agents exist, what they do, how they behave, and what risks they introduce. About the role Onyx is looking for an AI Researcher to help advance the core intelligence behind our AI security platform. In this role, you will explore emerging risks, behaviors, and capabilities of modern AI systems and translate your research into practical techniques that improve how organizations understand, monitor, and secure AI usage. You will work closely with applied AI engineers, security researchers, and product teams to turn cutting-edge research into real product impact. What you will do Conduct research on AI systems, agent architectures, and emerging security risks in enterprise AI environments. Design novel approaches for analyzing, evaluating, and monitoring AI agent behavior. Develop prototypes and research tools to test hypotheses and validate new techniques. Collaborate with applied AI engineers to translate research into scalable product capabilities. Stay up to date with the latest research in AI safety, AI security, and model behavior. Contribute to internal research documentation and, when appropriate, publish research to the broader community. Requirements What you will bring 3+ years of experience in AI research, machine learning research, or related fields. Strong background in machine learning, LLMs, or AI systems. Proficiency in Python and ML frameworks such as PyTorch, TensorFlow, or similar. Experience designing experiments and evaluating model behavior. Ability to translate research insights into practical engineering directions. Strong analytical thinking and communication skills. Nice to Have Experience researching AI security, AI safety, or adversarial ML. Familiarity with agent architectures, tool use, and multi-agent systems. Experience evaluating LLM behavior, prompt injection risks, or model misuse. Publications in top conferences (NeurIPS, ICML, ICLR, etc.) or applied research in industry. Background working with production AI systems or AI platforms.