remote
AI Security Engineer - World Wide Technology
Security Engineer
Design and secure large‑language‑model applications, implementing adversarial defenses, authentication, and monitoring. Leverage Python and LLM orchestration frameworks (LangChain, LlamaIndex, CrewAI) to protect AI services in production.
About the role
Key Responsibilities
- Architect and harden LLM‑based solutions, including Retrieval‑Augmented Generation, tool integration, and agent orchestration using frameworks such as LangChain, LlamaIndex, or CrewAI.
- Develop and apply adversarial ML defenses—prompt injection mitigation, data‑poisoning detection, model inversion prevention, and evasion‑attack safeguards.
- Implement robust identity and secret management: service‑account hygiene, secrets rotation, and OAuth/OIDC for machine‑to‑machine authentication.
- Establish continuous security monitoring for AI workloads, creating behavioral baselines, output anomaly detection, and comprehensive audit logging.
- Collaborate with ML engineers to embed security best practices throughout the model lifecycle, from data ingestion to deployment.
Requirements
- 4–7 years of software or machine‑learning engineering experience, with at least 2–3 years focused on AI/ML security.
- Strong proficiency in Python and hands‑on experience building LLM applications.
- Practical knowledge of adversarial techniques and mitigation strategies for large language models.
- Experience with OAuth/OIDC, service‑account management, and secret‑rotation processes.
- Ability to design monitoring pipelines that detect anomalous model behavior and maintain audit trails.