Lead Engineer, Durable AI Agent Runtime

ego AI (YC W24); San Francisco (CA); Hybrid; USD 150,000 - 200,000; Full time

Job summary

Ego AI, a YC-backed applied AI research lab in San Francisco, is hiring for a Member of Technical Staff to own the agent runtime. The role focuses on building a durable, extensible agent harness with a robust plan-tool-observe loop, per-user durable agents, and strong session management.

Hybrid work supported. You will design a per-user memory-enabled agent system, implement a tool registry, and ensure reliable execution from plan through tool calls, with emphasis on safety, observability, and

Qualifications

• 3+ years building backend or infrastructure systems with production ownership
• Hands-on experience building LLM agent systems: tool-calling loops, defensive handling of model output, and budgets
• Deep grasp of durable execution: checkpointing, idempotency, outbox/step-log patterns
• Deliberate context-management strategy for long-horizon tasks and trade-offs
• Strong API and data-model design; persistence behind clean seams
• Clear written communication and architecture documentation

Responsibilities

• Own the agent harness and core decision loop (plan → tool call → observe → repeat) as a maintainable state machine
• Build checkpointed, resumable sessions with durable logs and idempotent tooling
• Develop compaction, running summaries, and retrieval over older state for long-horizon tasks
• Design a tool registry with name, JSON schema, and handler; ensure safe schema validation
• Ship deployable systems with local bring-up, persistence interfaces (Postgres/SQLite/Cloudflare storage), and per-user durable agents
• Support engineers by reviewing designs and extending the harness

Skills

Backend systems

LLM agent systems

Durable execution

Context management

API & data-model design

Clear documentation

Tools

Cloudflare Durable Objects

Postgres/SQLite

Job description

Ego AI, a YC-backed applied AI research lab in San Francisco, is hiring for a Member of Technical Staff to own the agent runtime. The role focuses on building a durable, extensible agent harness with a robust plan-tool-observe loop, per-user durable agents, and strong session management.

Hybrid work supported. You will design a per-user memory-enabled agent system, implement a tool registry, and ensure reliable execution from plan through tool calls, with emphasis on safety, observability, and