Senior AI Engineer, Security Infrastructure
About this role
Engineers who thrive at the intersection of applied research, offensive security, and production systems engineering will find a natural home tackling complex adversarial challenges at Air. The team builds an AI-native enterprise readiness platform centered around agentic capabilities deployed across government and industrial sectors. As the AI agent Ace gains access to more powerful tools and data, the organization needs a dedicated specialist to secure these deployments against novel vulnerabilities and failure modes. You will work closely with engineers developing agent runtimes, evaluation infrastructure, and production systems to turn research insights into durable security architecture.
The role exists to ensure that powerful AI agents operate safely in high-stakes, adversarial environments by developing robust mitigations, sandboxing mechanisms, and automated evaluations. Ideal candidates possess a strong background in systems security, machine learning infrastructure, or adversarial testing, alongside the ability to translate complex threat models into concrete engineering controls. If you enjoy diving deep into distributed systems, investigating production failures, and establishing security primitives for next-generation AI platforms, this position offers a compelling technical scope.
What you will do
- Research and develop new approaches to AI red teaming, adversarial testing, security evaluation, and robust inference.
- Threat model agentic AI architectures to identify trust boundaries, attack surfaces, and potential failure modes.
- Design adversarial evaluations targeting prompt injection, indirect prompt injection, tool abuse, and privilege escalation.
- Build automated security evaluation and regression frameworks that continuously test agents, models, and tools.
- Translate successful attacks and research findings into production mitigations and reusable security controls.
- Design secure execution environments, sandboxing, and isolation mechanisms for handling untrusted workloads.
- Own and improve production infrastructure across Kubernetes, AWS, networking, storage, and compute.
What you need
- U.S. Citizenship is required.
- Bring 5+ years of experience building production software, backend infrastructure, distributed systems, security systems, or AI/ML infrastructure.
- Hold a degree in Computer Science, Computer Engineering, Cybersecurity, Data Science, or a related technical field, or possess equivalent practical experience.
- Demonstrate hands-on experience in AI red teaming, adversarial machine learning, offensive security, or systems security research.
- Maintain a strong understanding of how modern LLM and agentic systems operate, including inference, tool use, and multi-step execution.
- Possess strong programming experience in Python and the ability to build production-quality software.
- Have experience operating production services on Kubernetes and cloud platforms such as AWS, GCP, or Azure.
Nice to have
- Experience designing or securing AI agent runtimes and tool-execution environments.
- Experience building secure code execution sandboxes, container isolation, or microVMs.
- Experience with cloud security, infrastructure hardening, IAM, secrets management, and zero-trust architectures.
- Familiarity with AI security frameworks and threat taxonomies such as MITRE ATLAS, OWASP guidance, or the NIST AI Risk Management Framework.
- Experience working in government, defense, or other high-security environments.
Skills and tech stack
Questions candidates ask
Where is this role located?
This role is a full-time position located out of the office in Pittsburgh, PA or Arlington, VA, with remote options available for those outside of those cities. The position may require up to 25% travel.
What citizenship is required for the position?
U.S. Citizenship is required for this role due to the nature of the work supporting government and defense capabilities.
What will I be working on regarding AI security?
You will identify how agentic systems can fail or be exploited, develop new approaches for detecting and mitigating those failures, and build the infrastructure necessary to deploy capable AI agents securely in adversarial environments. This includes threat modeling, building automated security evaluations, and designing secure execution environments.