Software Engineer, Agent Harnessing

Scout AI•Sunnyvale, CA

About The Position

As a Software Engineer on Scout AI’s Agent Harnessing team, you will build the core architecture that enables AI agents to interact with and control complex, real-world physical systems. The platform spans agent orchestration, tool integration, low-latency communication layers, and execution environments, and is designed for Scout AI engineers to extend and depend on. The work includes developing the agent harness used across Scout AI’s agentic stack, connecting AI models to physical systems, building simulation environments for agent testing and training, and developing data pipelines and model fine-tuning that support agent training. You will also work directly with warfighters to translate capability requests into production-ready agentic systems.

Requirements

At least 5 years of full-stack engineering experience; startup or high-ownership environment experience is preferred.
Experience building complex agentic products or infrastructure, such as agent loops, orchestration systems, or tool-use frameworks.
Strong opinions on model and agent behavior.
Systems-oriented full-stack engineering experience, including hands-on integration of software with physical hardware or low-level systems.
Strong software engineering fundamentals in architecture, distributed systems, and real-time pipelines, and an interest in shipping production systems.
Must be a U.S. Person due to required access to U.S. export-controlled information or facilities.
Passion for the defense of the United States and its allies.

Nice To Haves

Exposure to simulation environments, game engines such as Unreal, Unity, or Isaac Sim, or robotics middleware such as ROS/ROS2.
Startup or high-ownership environment experience.

Responsibilities

Design and advance the agent harness—including the agent loop, tools, prompts, execution environment, and capabilities—used across Scout AI’s agentic stack.
Integrate full-stack software with physical hardware systems and AI agents, ensuring robust, low-latency communication across complex pipelines.
Build simulation environments and video-game-like frameworks that allow agents to be tested and trained before physical deployment.
Own the connectors between AI models and real-world systems that translate model output into physical action.
Fine-tune frontier AI models and build data acquisition pipelines that feed the training infrastructure powering the agents.
Work shoulder-to-shoulder with warfighters to turn capability requests into production-ready agentic systems.