AI Agent Architect, Customer Experience

Airtable Airtable · Enterprise · United States · Remote · Customer Support

The AI Agent Architect will own the technical foundation for Airtable's AI-native customer support experience, focusing on designing and optimizing AI agent reasoning, retrieval, decision-making, and action execution. This involves architecting knowledge systems, decision logic, and guardrails for reliable and scalable AI resolution, with a strong emphasis on production systems and LLM fluency.

What you'd actually do

  1. Own Agent retrieval accuracy and relevance. Architect the knowledge systems that enable AI agents to surface the right answer on the first try. Measure and improve retrieval precision, contextual relevance, and hallucination rates.
  2. Drive automated resolution rates. Build the decision frameworks that allow agents to take confident actions. What APIs do agents need to access? When can they make account modifications? You're accountable for encoding business logic into auditable, predictable systems that resolve issues without human intervention.
  3. Manage AI safety and trust. Establish the guardrails that keep resolution rates high while failure rates stay low. You're responsible for what the agent _doesn't_ do wrong: edge cases caught, prompt injection blocked, unintended behaviors prevented.
  4. Own the feedback loop. Monitor the observability layer that turns agent behavior into actionable insights. Instrument retrieval accuracy, action success rates, and failure patterns. Use this data to drive measurable week-over-week improvements in agent performance.
  5. Continuously improve agent quality. Develop and maintain the prompt architecture that governs how agents reason and respond. Build systematic approaches to versioning, A/B testing, and performance evaluation, measuring consistency, accuracy, and adaptability across scenarios.

Skills

Required

  • Deep understanding of LLM reasoning and failure modes
  • Hands-on experience with AI agent architectures
  • Experience building or contributing to AI-powered systems in production
  • Proficiency in RAG architectures, prompt engineering, chain-of-thought reasoning, and agent frameworks
  • System design for data flows, state management, error handling, and edge cases
  • Scripting, API integration, database querying, and prototyping
  • System instrumentation, log analysis, and data-driven diagnosis
  • Ability to explain complex AI systems to non-technical stakeholders
  • Technical documentation and system specification writing
  • Cross-functional collaboration with Engineering, Product, and Operations

Nice to have

  • Solutions architecture
  • Platform engineering
  • Technical program management

What the JD emphasized

  • AI agents
  • LLMs
  • production systems
  • guardrails
  • retrieval accuracy
  • hallucination rates
  • automated resolution rates
  • agent safety
  • trust
  • observability layer
  • agent performance
  • prompt architecture
  • performance evaluation

Other signals

  • AI agents
  • LLMs
  • production systems
  • customer support