Software Engineer, Agent Platform

Anduril's Lattice software platform integrates together many sensors into a single cohesive view of the world, providing needed context for our users. Anduril's Frontier AI team builds edge-compatible, generative AI systems into the Lattice software platform to provide features and products that improve autonomy and reduce cognitive burden on the warfighter.

Specific applications include but are not limited to automating mission planning, battle-space understanding, voice-control of assets, and enabling higher-levels of autonomy.

Frontier AI is looking for a backend software engineer to own and evolve our internal LLM agent framework. This role sits at the intersection of backend infrastructure, applied AI, agent architecture, model post-training, and evaluation tooling. You will build the platform that enables teams across Anduril to develop, evaluate, and deploy reliable LLM agents in mission-critical environments.

- Own Anduril's internal LLM agent framework, including core abstractions, runtime architecture, developer experience, integrations, and reliability.

- Support multiple business lines building LLM agents by providing new framework capabilities, implementation guidance, architectural reviews, and best-practice patterns.

- Partner with machine learning teams to make model post-training workflows easy to integrate, ranging from supervised fine-tuning to offline RL, online RL, and environment-driven agent improvement.

- Design tooling that supports modern agent patterns, including structured tool calling, filesystem-using agents, memory and retrieval, planning loops, subagents, agent graphs, and human-in-the-loop workflows.

- Work with partner teams to define comprehensive evaluation suites that measure task success, tool-call correctness, trajectory quality, robustness, regressions, and deployment readiness.

- Stay current on emerging agent architecture and evaluation trends, and make pragmatic decisions about which techniques should or should not be adopted internally.

- Strong backend engineering experience building production-quality platforms, frameworks, APIs, or infrastructure used by other engineers.

- Deep expertise in LLM agent framework design, including the tradeoffs between different orchestration patterns such as linear agents, graph-based agents, multi-agent systems, planner/executor loops, and tool-heavy agents.

- Experience designing agent evaluation paradigms, including trajectory evaluations, LLM-as-judge workflows, task-success metrics, tool-call correctness checks, rubric-based qualitative grading, adversarial scenario testing, regression eval suites, and human-in-the-loop review.

- Familiarity with model post-training workflows such as SFT, preference tuning, reinforcement learning, and environment-based agent training.

- Strong judgment around reliability, observability, debugging, and safety for LLM applications deployed in high-stakes settings.

- Ability to work directly with partner teams, understand ambiguous product needs, and turn them into reusable platform capabilities.

- Experience with agent frameworks like Langchain Deepagents, Claude SDK, etc.

- Experience building evaluation platforms, simulation environments, benchmark suites, or agent test harnesses.

- Experience with Kubernetes, Docker, distributed systems, workflow orchestration, or ML infrastructure.

- Familiarity with defense, robotics, command-and-control systems, autonomy, or operational planning domains.