- San Francisco, California, United States
- Full-time · Senior
- $234K–$384K per year
- 5+ years
Join Cisco’s AI Threat Intelligence and Security Research team to design the frontline defense for enterprise AgenticOps! Powering over 80% of the world’s internet traffic, Cisco provides the global foundation where autonomous AI agents, enterprise software, and cloud networks converge. Our best-in-class research pod investigates emerging threat vectors targeting generative foundation AI models, agentic workflows, and distributed AI platforms and supply chains, then translates our research into practical protections—including red teaming, security tools, threat detections, provenance capabilities, runtime guardrails, and product innovations.
Compensation
The starting salary range posted for this position is $256,900.00 to $333,400.00 and reflects the projected salary range for new hires in this position in U.S. and/or Canada locations, not including incentive compensation*, equity, or benefits. The applicable full salary ranges for this position, by specific state, are listed below. New York City Metro Area: $256,900.00 - $383,500.00. Non-Metro New York state & Washington state: $234,400.00 - $341,100.00.
Responsibilities
- Lead research on adversarial charges targeting generative AI and agentic systems, including prompt injection, jail-breaking, agent goal drift or misalignment, backdoors, model data poisoning, model evasion, and malicious tool-calling exploitation.
- Architect robust agentic security frameworks and defense-in-depth guardrails across open agent protocols (e.g., MCP), multi-agent orchestration, and autonomous execution environments.
- Design, evaluate, and deploy automated threat detection models, behavioral telemetry signatures, and real-time inference guardrails into enterprise production platforms.
- Design, perform, evaluate, and iterate on sophisticated adversarial red-teaming and safety benchmarking across foundation models, fine-tuned LLMs, and Retrieval-Augmented Generation (RAG) pipelines, mapping findings to the MITRE ATLAS matrix.
