Staff Software Engineer, AI Foundations (agent Optimization)

Temporal Temporal · Enterprise · United States · AI SDK

Staff Software Engineer focused on optimizing AI agents for Temporal users, aiming to improve token spend, response time, and result quality through techniques like model routing, multi-agent architecture, cache policy, and context management.

What you'd actually do

  1. Work as a software engineer
  2. Maintain and expand a deep understanding of agentic coding
  3. Design and build agentic coding systems that we can trust to deliver high-quality outputs
  4. Develop a deep understanding of AI application development patterns and techniques, including emerging approaches and architectures.
  5. Take end-to-end ownership of new features, working with other teams to deliver exceptional reliability and a great developer experience.

Skills

Required

  • 8+ years of professional work experience
  • Proven experience in optimizing AI systems using both machine learning and systems techniques
  • Experience and passion for harnessing generative AI, particularly for agents and coding.
  • Python
  • TypeScript
  • Java
  • Go

Nice to have

  • A deep appreciation for how to use AI to improve quality, not only to increase quantity.
  • A sense of taste in code and software development practice. Your approach should be opinionated and thoughtful, but not dogmatic.

What the JD emphasized

  • direct experience with this problem domain
  • optimizing AI systems using both machine learning and systems techniques
  • harnessing generative AI, particularly for agents and coding
  • use AI to improve quality, not only to increase quantity

Other signals

  • design tools and mechanisms to help Temporal users build agents that are optimized for token spend and response time while maintaining result quality
  • model routing is a first step, but represents the tip of the iceberg of techniques that we can apply
  • multi-agent architecture, cache policy, and context management are all relevant
  • direct experience with this problem domain