Agent Systems Engineer — Build Adaptive, Reliable AI Agents

Adaption Labs; San Francisco (CA); Hybrid; USD 180,000 - 240,000; Full time

Benefits offered by this job

Flexible work

Adaption Passport

Lunch Stipend: Weekly meal allowance

Well-Being: Medical benefits and PTO

Job summary

Adaption Labs is building agent systems at the core of our product in the Bay Area. You'll design and implement architectures that turn customer goals into reliable, multi-step executions across real tools and services, with real constraints and recovery from failures.

You will own planning, tool use, memory, and observability for agents, and develop evaluations to measure performance while balancing quality, latency, and cost.

Qualifications

• 5+ years building production ML or backend systems, including taking LLM or agent applications from prototype to production.
• Strong understanding of agent design: planning, reasoning, tool use, orchestration, and memory.
• Experience building rigorous evaluation systems, plus execution tracing and observability for agents, with a focus on reproducibility.
• Familiarity with the OpenAI Responses API, MCP, and server- versus client-side execution.

Responsibilities

• Design agent architectures for planning, reasoning, tool use, memory, and integration with external systems and data.
• Improve reliability on long-running, multi-step tasks, including failure recovery.
• Build the loops that let agents improve with real use, so performance compounds instead of staying frozen.
• Develop evaluations that measure agent performance and resist being gamed.
• Make practical tradeoffs between quality, latency, cost, and complexity.

Skills

Agent design

Planning & reasoning

Observability

Adaptability

Tools

OpenAI Responses API

MCP

Job description

Adaption Labs is building agent systems at the core of our product in the Bay Area. You'll design and implement architectures that turn customer goals into reliable, multi-step executions across real tools and services, with real constraints and recovery from failures.

You will own planning, tool use, memory, and observability for agents, and develop evaluations to measure performance while balancing quality, latency, and cost.