San Francisco, California30+ days ago
Whiteboard architecture for systems running autonomous agents continuously, where every architectural decision impacts token costs, execution latency, and quality; Review critical-path code in our agentic orchestration layer, not to micromanage, but because decisions here ripple to millions of requests; Solve distributed systems problems that arise when coordinating multiple LLM-based agents with bounded autonomy, managing context windows, and optimizing state persistence; Unblock engineers by doing hands-on technical work: refactoring a bottleneck, designing a new subsystem, debugging a production incident that requires deep systems thinking. The problems we're solving, perpetual agent coordination, cost-quality tradeoffs in LLM orchestration, distributed state management in autonomous systems, aren't taught in universities or discussed in most engineering organizations.