Confirmed on the employer's own hiring board on Nov 19, 2026. First seen by Alion on Nov 19, 2026.
About the Role
Build and deploy production-grade agentic AI systems for high-impact government missions. This role combines hands-on engineering with direct collaboration with government partners to turn mission needs into operational systems.
What You'll Do
Build and deploy LLM-powered agentic systems, including multi-agent, retrieval-augmented generation, and human-in-the-loop solutions.
Work with government partners in field environments to translate mission needs into technical systems.
Design and maintain data curation, evaluation, and machine learning infrastructure pipelines.
Own systems through architecture, implementation, deployment, reliability, and iteration.
Develop new AI systems that connect mission data and models to operational autonomy.
What We're Looking For
More than two years of professional experience in software, AI/ML, or solutions engineering, with strong end-to-end engineering ownership.
Strong Python and TypeScript skills, with experience building and shipping customer-facing AI agents into production.
Deep expertise in model pre-training or post-training, or in RAG and retrieval systems.
Experience building data pipelines, labeling workflows, or ML infrastructure.
Ability to present technical work to customers and operate independently in ambiguous, partner-facing settings.
Experience in high-security or high-reliability environments; government, national security, or defense technology experience is helpful.
Familiarity with reinforcement learning workflows or simulation environments is helpful.
A technical degree and an active TS/SCI security clearance are required. Must be authorized to work in the United States without visa sponsorship and willing to travel up to 20% to partner sites.
Compensation & Benefits
Base salary range: $160,000 to $250,000 annually. Visa sponsorship is not available.
Location
Hybrid role based in Washington, United States, with travel to government partner sites as needed.

