The opportunity
At Klaviyo, we believe the future of software lies not only in tools that help people work more efficiently, but in intelligent systems that can take action, learn from outcomes, and improve customer experiences over time.
What you'll do
Design and build reliable, scalable backend systems and APIs that power AI-driven customer experiences.
Build and evolve production agentic systems involving tool use,: orchestration, context management, retrieval, structured outputs, and multi-step workflows.
Solve hard applied AI problems across retrieval and RAG, grounding, model: selection, hallucination mitigation, and deciding when deterministic or hybrid approaches are better than LLMs.
Develop evaluation strategies using representative datasets, automated and: human evaluation, regression testing, qualitative error analysis, and production signals.
Improve production reliability through guardrails, fallbacks, retries,: observability, monitoring, and thoughtful failure handling.
Build asynchronous and distributed processing workflows that support high-volume AI workloads.
What they're looking for
- You have 5-7 years of professional software engineering experience , with a: strong foundation in backend systems and distributed applications.
- You have hands-on experience building and deploying generative AI or agentic: AI applications into production for real users.
- You are proficient in Python and modern backend frameworks such as FastAPI or Django.
- You have experience with asynchronous processing and distributed task or: event systems such as Celery, Kafka, SQS, RabbitMQ, or Redis.
- You have strong knowledge of databases, data modeling, APIs, and persistence: patterns used in production systems.
- You are comfortable working in cloud-native environments with technologies: such as AWS, containers, Kubernetes, infrastructure automation, or CI/CD systems.
- You can reason clearly about tradeoffs among quality, latency, cost,: reliability, and implementation complexity.
- You are able to operate independently in ambiguous environments, break down: loosely defined problems, and make pragmatic decisions about what to build first.
