AI Agent Systems Software Engineer, Senior

Booz Allen Hamilton•Mclean, VA

1d•Remote

About The Position

As an AI software engineer, you've moved past building simple LLM wrappers and into the harder problem: building agents that interact with structured, deterministic systems, and produce results that are reliable enough to act on. You understand that the real challenge isn't getting a language model to generate text. It's getting an autonomous system to call the right tools in the right order, translate between natural language and formal representations, and know when it has made an error. We are building a platform where AI agents serve as the bridge between human intent and formal computational systems, and we need someone like you to design and build that bridge. On this team, you'll own the AI agent layer that sits between operators and the platform's core reasoning and modeling capabilities. You'll design agents that can interpret loosely specified requests, discover relevant computational assets, translate requirements into structured inputs that formal systems can evaluate, and explain results back to users in language they can act on. When the underlying systems reject an input or return a counterexample, your agents need to localize the error, suggest corrections, and help users iterate quickly. You'll work with LLMs, tool-use frameworks, and retrieval systems, but always in service of a system where the final answers must be formally grounded, not probabilistically generated. You'll join a small, elite team where everyone operates at the technical frontier. You'll be the person on the team who makes the system accessible, designing agents that let operators interact with formal reasoning and computational modeling through natural language rather than code. You'll build in a rapid development environment with AI-native tooling, measuring success by whether operators trust and have confidence in your agents. Join us. The world can't wait.

Requirements

5+ years of experience designing and implementing software systems
2+ years of experience on applications involving large language models, autonomous agents, or natural language processing
Experience building agentic systems that use tool calling, multi-step reasoning, or retrieval-augmented generation to interact with external APIs or structured data sources
Experience with LLM orchestration frameworks such as LangGraph, LangChain, or agent-coordination tooling
Ability to work across the boundary between probabilistic AI systems and deterministic software, ensuring that LLM outputs conform to the expectations of formal or structured downstream systems
Ability to obtain a Secret clearance
Bachelor's degree in computer science, artificial intelligence, or computational linguistics

Nice To Haves

Experience translating natural language specifications into formal or semi-formal representations such as constraint sets, typed schemas, or domain-specific intermediate languages
Experience building human-in-the-loop AI workflows where the system proposes actions and a human approves, modifies, or rejects them
Experience deploying or fine-tuning language models in secure or air-gapped environments for defense, intelligence, or national security customers
TS/SCI clearance with a polygraph
Master's degree in computer science or artificial intelligence preferred
Doctorate degree in computer science or artificial intelligence a plus

Responsibilities

Own the AI agent layer that sits between operators and the platform's core reasoning and modeling capabilities.
Design agents that can interpret loosely specified requests, discover relevant computational assets, translate requirements into structured inputs that formal systems can evaluate, and explain results back to users in language they can act on.
Localize errors, suggest corrections, and help users iterate quickly when underlying systems reject an input or return a counterexample.
Work with LLMs, tool-use frameworks, and retrieval systems to ensure final answers are formally grounded.
Design agents that let operators interact with formal reasoning and computational modeling through natural language rather than code.
Build in a rapid development environment with AI-native tooling.