Ahmad Al-Dahle. Chief Technology Officer at Airbnb, where he leads the company's AI-native engineering shift. Previously built and led the Generative AI function at Meta, overseeing the Llama foundation model series. At Airbnb, his teams built Everest, an internal knowledge graph that makes codebase context queryable by LLMs, enabling generalist engineers to work in specialist product areas. The company reports 60% of new code is AI-authored, 80% more features shipped year-over-year, and AI agents now resolve roughly half of all support tickets. Al-Dahle runs separate eval suites per use case, sampled from production queries, as the mechanism for deciding which ticket types are safe to hand to agents.
Operating themes
- Knowledge graph as expertise multiplier A queryable codebase context graph compresses domain onboarding for generalist engineers, enabling faster feature development without specialist bottlenecks.
- Coverage by deliberate exclusion Solving 50% of support tickets with clear criteria for which cases AI handles produces better outcomes than automating everything and failing on high-stakes queries.
- Per-use-case measurement Each AI deployment needs its own eval suite sampled from production queries, not a shared generic benchmark.