About the RoleWe're hiring software engineers for the Model Behavior team to help shape how Perplexity's AI products behave: the style of their responses, and the way they use tools, skills, and memory. The team designs prompt and context engineering strategies to deliver high-quality user experiences across multiple domains and models.
The ideal candidate for this role has a strong software engineering background, and an analytical, experiment-driven approach to solving challenging problems. You'll work on context and prompt engineering to shape model behavior and style, and to guide how models use tools, skills, and memory across our products.
What you'll do- Context Engineering: Design, test, and optimize the prompts, skills, tools, and memory that shape Perplexity responses across products, features, and use cases. Build self-improvement loops to steer the prompt, improve tool/skill use, and improve the ability to draw on memory.
- Model Releases: Help experiment with and release new models.
- Research & Analysis: Identify inconsistencies and failure modes in model outputs through well-designed research projects, for both internal and production systems.
- Knowledge Sharing: Help engineers across teams build intuition for prompt design and context engineering best practices.
- Staying Current: Track the latest prompting, context engineering, and alignment techniques from industry and academia, and bring the best ideas back to the team.
What We're Looking ForRequired- 2 to 10+ years of experience in software engineering or research.
- Strong background in software engineering fundamentals, and a technical understanding of LLM-driven and agentic systems.
- Experience shaping LLM behavior through prompts, tool and skill design, or memory systems.
- Strong written and verbal communication skills, particularly in explaining complex concepts to diverse stakeholders.
Nice to have- Recent experience working on modern LLM-driven products.
- Experience working across teams or with external partners.
- Experience designing evaluations or benchmarks for AI systems.
