Anduril Industries
AI Systems Engineer
About this role
Anduril Industries seeks an Applied LLM Systems Engineer to design and operate production AI systems that streamline technical documentation workflows. You'll build LLM-assisted authoring, publishing automation, and multi-agent coordination tools while managing cost, latency, and safety in a defense technology environment.
What you'll do
- Build production-grade LLM applications for documentation and knowledge-work processes
- Design multi-step workflow orchestration that safely coordinates models, tools, and deterministic services
- Optimize context management, token usage, and cost across LLM systems at scale
- Create evaluation frameworks and regression testing for prompts, models, and end-to-end workflows
- Establish observability, auditability, and rollback mechanisms for AI-assisted workflows
- Partner with technical writers and engineers to identify high-value AI applications and maintain human oversight where needed
What they're looking for
- Python and modern API development
- Production LLM system design and optimization
- Workflow orchestration and multi-agent coordination
- RAG, structured outputs, and tool-calling patterns
- Evaluation and testing frameworks for nondeterministic systems
- Software architecture and security boundaries
- Cloud infrastructure and CI/CD
- Cost and latency optimization
Opens the application — the Jobs AI extension fills it for you. Set up autofill
Opens the official application on the employer’s site. No login required.
Anduril Industries
Anduril Industries builds autonomous defense systems including underwater vehicles, unmanned aircraft, and electronic warfare platforms for the Department of Defense. The company is hiring across mechanical engineering, mission operations, software development, technical leadership, and advanced manufacturing roles to support the design, deployment, and production of these mission-critical systems.
- Website
- anduril.com
Likely interview questions
- Describe a production AI system you've delivered end-to-end—what were the key challenges in making it reliable and cost-effective at scale?
- How have you approached evaluation and regression testing for systems where LLM outputs are nondeterministic?