Staff Engineer, Applied AI (Customer Agent)New$264K

The opportunity

At Klaviyo, we value the unique backgrounds, experiences and perspectives each Klaviyo (we call ourselves Klaviyos) brings to our workplace each and every day. We believe everyone deserves a fair shot at success and appreciate the experiences each person brings beyond the traditional job requirements.

What you'll do

  • Set the technical and product direction for a core area of Customer Agent,: from architecture through production, and contribute to the direction of the platform overall.

  • Take on the hardest technical problems in your area: orchestration, evaluation, guardrails, memory & context, failure handling for non-deterministic systems.

  • Write and ship code, review designs, and make the hard calls in your area: when there’s no obvious right answer.

  • Define patterns that other engineers adopt, and mentor through design reviews rather than direct management.

  • Partner with product, design, data science, and go-to-market leaders to shape: what gets built, not just how it’s built.

  • Evaluate emerging AI technologies and bring the ones worth betting on into production.

What they're looking for

  • You've built, scaled, and supported production agentic systems, not just: prototyped them: real orchestration, evals, guardrails, and non-deterministic failure handling. You've owned your part of one through launch, incidents, and the operational reality after the demo.
  • You've learned the hard lessons that only come from running agents in: production: securing them against misuse and abuse, and the operational scar tissue of keeping a non-deterministic system reliable as usage grows.
  • You've owned technical and product direction for a significant area before,: not just your own component, and you can point to something you designed and championed that shipped and that other engineers built on.
  • You're hands-on: writing code, reviewing your own and others’ architecture decisions, and debugging production issues yourself, not delegating the hardest problems away.