Backblaze External Website - Sr. AI Security Engineer
Requirements
• Experience securing internal AI platforms or developer-facing AI tools • internal AI platforms or developer-facing AI tools • Background in detection engineering, threat hunting, or adversarial testing • detection engineering, threat hunting, or adversarial testing • Familiarity with agent frameworks (e.g., LangChain, LlamaIndex) • Experience mentoring engineers and guiding secure design • At this point, we hope you're feeling excited about the job description you're reading. Even if you don't meet every requirement, we still encourage you to apply. Learning, developing, and growing are key parts of our culture. We're eager to meet people who believe in our mission and can contribute to our team in various ways. We want people to feel comfortable expressing their true selves and to come, stay, and do their best work here.
Responsibilities
• Agentic AI Safeguards • Architect and implement guardrails for tool-using AI systems, including: • guardrails for tool-using AI systems • Tool access controls and allowlists • Context and memory isolation • Step-level validation of agent actions • OWASP Agentic AI Top 10 • Runtime Security Controls • Build enforcement mechanisms that govern AI behavior at execution time: • enforcement mechanisms • Interceptors, proxies, or middleware for tool/API calls • Policy decision and enforcement layers • Rate limits, execution bounds, and kill-switches • Prevent unsafe or unauthorized actions initiated by AI systems • Non-Human Identity (NHI) • Design and implement identity and access controls for agents and automation, including: • identity and access controls for agents and automation • Short-lived credentials and scoped permissions • Clear separation between human and non-human access • Strong binding of identity to task context and execution • Ensure all AI actions are attributable and auditable • attributable and auditable • Observability & Detection • Implement logging and tracing for AI activity: • logging and tracing for AI activity • Prompts, tool usage, and decision flows • Build detection capabilities using: • Behavioral baselining and anomaly detection techniques • Identify and alert on: • Abnormal tool usage • Suspicious prompt patterns • Unexpected data access • Threat Modeling (MAESTRO) • Perform agentic system threat modeling using MAESTRO, including: • agentic system threat modeling using MAESTRO • Mapping agent capabilities, trust boundaries, and attack paths • Modeling misuse and adversarial scenarios • Translate findings into practical safeguards and detection logic • practical safeguards and detection logic • Protect developers using AI tools by: • Preventing sensitive data exposure • Validating AI-generated code and actions • Constraining unsafe automation • Enable safe usage of AI-assisted development tools (e.g., Claude Code, Codex, Cursor) with: • Claude Code, Codex, Cursor • Security validation layers • Controlled prompting and output handling patterns • The Right Fit: • The Right Fit: • 7+ years in security engineering or backend systems • security engineering or backend systems • Proven experience designing and deploying security controls, such as: • designing and deploying security controls • Runtime enforcement layers (proxies, middleware, policy engines) • Identity and access systems, especially for non-human entities • Strong programming skills (Python preferred; Go, Java, or TypeScript a plus) • Experience using AI-assisted development tools such as Claude Code in real workflows, including understanding associated security risks and safeguards • AI-assisted development tools such as Claude Code • Experience with: • Logging, monitoring, and detection systems • Building or securing API/service interactions • Practical familiarity with: • Agentic AI systems or tool-integrated LLM workflows • OWASP guidance for AI/agent risks • Practitioner Knowledge • Experience applying (not just referencing): • OWASP Agentic AI / LLM risk guidance • NIST AI RMF concepts in real systems • NIST AI RMF • CSA guidance on workload and machine identity • CSA guidance • Strong understanding of: • Zero Trust for non-human identities • Secrets management and credential scoping • Observability tooling (e.g., OpenTelemetry, ELK)
Apply in one click
Upload My Resume
Drop here or click to browse · Tap to choose · PDF, DOCX, DOC, RTF, TXT