Applied LLM Systems Engineer

Anduril Anduril · Defense · Santa Ana, CA · Programs : Deployments : Technical Publications

This role focuses on building and operating production AI systems, specifically LLM applications for technical documentation workflows. It involves designing multi-step agentic systems, optimizing for cost and latency, and implementing robust evaluation and observability frameworks. The role requires experience delivering production AI systems and managing probabilistic systems in an enterprise environment.

What you'd actually do

  1. Build production-grade LLM applications for documentation and knowledge-work workflows.
  2. Design orchestration for multi-step workflows that coordinate models, tools, and deterministic services safely.
  3. Optimize context management, token usage, caching, and workflow design for cost, latency, and task success.
  4. Build retrieval, structured-output, and tool-integrated systems that safely interact with enterprise content, source control, issue tracking, and documentation systems.
  5. Design evaluation frameworks and regression testing for prompts, models, tools, retrieval pipelines, and end-to-end workflows.

Skills

Required

  • Python
  • modern API, data, and service-development practices
  • software architecture
  • security boundaries
  • identity and access management
  • data protection

Nice to have

  • multi-step orchestration or coordination systems
  • retrieval-augmented generation
  • structured outputs
  • tool calling
  • production LLM workflow patterns
  • defense, aerospace, manufacturing, robotics, autonomy, aviation, or another regulated environment
  • technical documentation
  • structured authoring
  • content management
  • containerized services
  • cloud infrastructure
  • CI/CD
  • queues
  • event-driven architectures
  • production monitoring

What the JD emphasized

  • production AI systems
  • measurable adoption by its intended users
  • optimizing cost, latency, and context usage
  • designing evaluation, observability, rollback, and failure-handling for nondeterministic systems

Other signals

  • production AI systems
  • LLM applications
  • multi-agent coordination