Radley James has partnered with an AI start-up building advanced AI systems for mission-critical and high-security environments. This is an opportunity to work on technology where reliability, reasoning and real-world performance genuinely matter.
The role will focus on building the core reasoning and orchestration layer behind the platform, working across multi-agent systems, long-horizon planning, tool use and agentic workflows within highly non-deterministic environments.
What You’ll Do
· Design and implement multi-agent architectures: supervisory agents, planning layers, session management, tool orchestration, and failure recovery across long-running autonomous workflows
· Build and refine the reasoning and decision-making systems that govern complex autonomous behaviour — strategic adaptivity, environmental awareness, and adversarial resilience
· Develop evaluation and observability infrastructure for agentic systems operating in real-world, non-deterministic environments
· Architect LLM integration patterns (structured generation, function calling, context management) with a deep understanding of model behaviour, failure modes, and capability boundaries
· Prototype aggressively — the correct answer to “can we do X” is a working proof-of concept by end of day, not a design doc by end of week
· Work across the stack as needed — Python, TypeScript, infrastructure — with primary ownership of the AI/reasoning layer
· Collaborate directly with the founder and field engineers to translate mission requirements into agent capabilities on compressed timescales
Requirements
1 - 6 years of experience as an AI systems engineer or an AI product engineer.
C. S. Degree from a reputable school (e. g. , Berkeley, Georgia Tech, MIT, UIUC, etc. ) can be a positive signal.
Proficient in Python or TypeScript.
Must be a US Citizen
able to obtain security clearance.
Willing to work on- site in Washington, DC (5 days/week).
Relocation support is available.