AI Security Research Engineer II

Position: AI Security Research Engineer II (Full Time) - (United States)
The application window is expected to close on: 07/15/2026

** Job posting may be removed earlier if the position is filled or if a sufficient number of applications are received** .

Plans, conducts, analyzes, and shares applied security research that advances how AI is built and used securely across Cisco, spanning AI-assisted development, AI features in products, and AI in internal operations. Combines AI/ML security research with hands-on engineering to threat-model AI systems, build guardrails and automated checks, and validate that AI-developed and AI-assisted software is at least as secure as traditionally developed software.

Serves as a security partner to engineering and product teams, and shares results that influence how teams ship AI safely and quickly. The role enables teams to adopt AI securely rather than acting as a gatekeeper.

What You'll Do:

- Executes straightforward, short-term AI security initiatives with guidance (e.g., one- to three-month deliverables such as a new guardrail, a Git Hub Action security check, or an evaluation harness).

- Influences a single feature or product team, helping them reach a secure outcome faster with AI, not despite it.

- Applies a working understanding of AI/ML security concepts and frameworks (OWASP LLM Top 10, OWASP Agentic Top 10, MITRE ATLAS, NIST AI RMF) and develops judgment on ethical and responsible-AI considerations.

- Supports AI-specific threat models for AI-native features (RAG, tool use, agentic workflows, MCP (Model Context Protocol) integrations), including risks such as prompt injection (direct and indirect), jailbreaks, data poisoning, model extraction, excessive agency, and supply-chain risk.

- Runs defined adversarial tests and evaluations using Cisco AI Defense (Cisco's AI security solution) alongside open-source tooling such as NVIDIA's Garak; documents findings, success rates, and remediation guidance.

- Builds or contributes to security tooling and guardrails (input/output filtering, automated checks, and release gates), delivered in CI/CD pipelines and agentic workflows under direction.

- Integrates AI security into existing Secure Development Lifecycle practices: threat modeling, SAST/DAST/SCA, dependency and supply-chain, secrets, and licensing.

- Supports data protection requirements for AI systems, including data classification, minimization, residency, and sensitive data handling.

- Supports human review, approval, and accountability controls for AI-assisted development and agentic workflows.

- Contributes to AI security standards, patterns, and security requirements, treating reference architectures as living documents that absorb emerging threats and engineering feedback.

- Develops and tests hypotheses about AI attack surfaces and control effectiveness; uses measurable evidence rather than manual assertions.

- Shares findings and demos with the team and partner groups with guidance; interacts with peers to understand cross-functional security needs.

- Contributes to reusable patterns, training material, office hours, and security champion enablement for engineering teams.

- Contributes to literature reviews, threat research write-ups, and presentations shared within the company; participates in internal and external AI security talks and discussions.

- Develops familiarity with the fast-moving AI/LLM platform, agent-framework, and attacker-technique landscape through hands-on experimentation.

Minimum Qualifications:

- Bachelor's + 2 years of related experience, or Master's + 0 years of related experience.

- Foundation in security engineering (application/product security, secure SDLC, threat modeling) and/or AI/ML engineering, with demonstrated adversarial thinking.

- Proficiency in Python.

Preferred Qualifications:

Varies based on team and business needs; in addition to

Minimum Qualifications .

- Hands-on exposure to LLM security: prompt injection defense, guardrails, insecure output handling, RAG and agentic-system risks.

- Familiarity with AI red-teaming/eval tooling, including Cisco AI Defense and an open-source framework such as NVIDIA's Garak.

- Experience with CI/CD, MCP (Model Context Protocol), agent-to-agent integration…