Senior AI Security Engineer

We are seeking a skilled software engineer with practical experience in machine learning (ML) and large language models (LLM) from a security standpoint. In this role, you will analyze findings from penetration tests and AI red-team exercises to identify root causes within code, models, data pipelines, or cloud configurations, and deliver robust, lasting fixes. Beyond remediation, you will design and implement defenses to prevent recurrence of similar issues, including guardrails around model endpoints, secure patterns for retrieval-augmented generation (RAG) and agents, and security checks integrated into CI/CD pipelines. Your work will span both internal AI products and client projects where remediation is required.

Key Responsibilities

Remediation:
- Triage and reproduce penetration test and AI red-team findings to determine root causes.
- Implement fixes across application code, APIs, model-serving layers, data pipelines, and cloud infrastructure.
- Develop regression tests for each resolved issue, including adversarial prompt tests within CI environments. Track remediation progress and communicate status to engineering leads and clients.

AI/LLM Defenses:
- Design and deploy guardrails for model endpoints, including input/output filtering, prompt-injection detection, personally identifiable information (PII) redaction, and policy enforcement.
- Establish secure patterns for RAG and agent frameworks, ensuring least-privilege access, sandboxed execution, human approval for high-risk actions, and tenant isolation.
- Secure the ML supply chain by managing model weights, datasets, and dependencies through signing, scanning, and provenance tracking.
- Implement logging, monitoring, and alerting for AI workloads, and develop incident response runbooks tailored to AI-specific events.

Secure Engineering:
- Conduct threat modeling for new AI features using frameworks such as OWASP Top 10 for LLM Applications, MITRE ATLAS, and NIST AI Risk Management Framework (RMF).
- Integrate security checks into CI/CD pipelines, including static application security testing (SAST), dependency and secret scanning, and model scanning.
- Review code and architecture for AI products and third-party AI services to ensure security compliance.
- Provide training to developers on secure AI development practices and create reusable security libraries and templates.

Required Qualifications
- 5 to 7 years of experience in software engineering, application security, or security engineering, with at least 2 years focused on security roles.
- Hands-on experience securing or developing at least one production LLM or ML application.
- Bachelor’s or Master’s degree in Computer Science, Software Engineering, Cybersecurity, Artificial Intelligence, or a related field, or equivalent practical experience.

Key Skills
- Proficiency with LLM and Generative AI engineering, including APIs such as OpenAI, Anthropic, Azure OpenAI, and Vertex AI, as well as open-weight models. Experience with RAG, embeddings, vector databases, and agent frameworks like LangChain or LlamaIndex.
- Expertise in AI threat mitigation, defending against prompt injection, jailbreaks, data leakage, poisoning, model extraction, and excessive agency, aligned with OWASP Top 10 for LLM Applications and MITRE ATLAS.
- Strong knowledge of secure-by-design AI, including input/output filtering, PII redaction, policy enforcement, least-privilege access, sandboxing, human approval workflows, and tenant isolation using tools such as NeMo Guardrails, Llama Guard, Lakera, or Presidio.
- Solid foundation in application security principles, including OWASP Top 10, authentication and authorization, secrets management, cryptography basics, and secure code review.
- Understanding of ML fundamentals and supply chain security, including model training, serving, evaluation, and securing model weights, datasets, and dependencies through signing and scanning (e.g., ModelScan).
- Experience with DevSecOps and CI/CD security integration, including SAST tools like Semgrep, dependency and secret scanners such as Snyk, Dependabot, Gitleaks, and automated adversarial prompt testing tools like Garak, PyRIT, or Promptfoo.
- Familiarity with containerization and cloud security, including Docker, Kubernetes, IAM, networking, and secure deployment of AI workloads on AWS, Azure, or GCP.

This role offers the opportunity to work at the intersection of software engineering, AI innovation, and cybersecurity, contributing to cutting-edge AI products and client solutions with a strong security focus.