Perplexity serves tens of millions of users daily with reliable, high-quality answers grounded in an LLM-first search engine and specialized data sources. The Answer Quality team ensures that our prompts, tools, search, and specialized datasets, combined with both frontier and in-house models, create the best possible experience for our users. As a Data Scientist/Engineer on this team, you will derive online signals from user interactions to bridge the gap between changes in answer quality and observed user behavior.
Responsibilities
- Discover and validate online signals from user interactions that serve as reliable proxies for true answer quality
- Design and implement novel online metrics to be tracked both in A/B testing and on product health dashboards, ensuring alignment with ground-truth evaluations
- Analyze experimental results to validate these metrics, ensuring they accurately predict user satisfaction and drive product decisions
- Build and maintain the data pipelines that calculate these metrics at scale, delivering actionable quality signals to Search, Product, and model training teams
- Communicate findings and bring clarity through close collaboration with Product and Search teams
- Operate in a small, high-impact team where your work directly shapes how Perplexity measures and improves Answer Quality
Qualifications
- MS in a technical field or equivalent experience
- 4+ years of experience working as a Data Scientist, Analytics Engineer, or related role
- Experience working on search, recommendation, or LLM-based products, with an emphasis on designing online metrics and analyzing A/B experiments
- Strong proficiency in Python and SQL (expected to write production-grade code)
- Deep knowledge of statistical analysis
- Experience with Business Intelligence (BI) tools for visualization and reporting
- Comfortable with agentic coding workflows and using AI-assisted development tools to iterate faster
Preferred Qualifications
- Proficiency with Apache Spark and Databricks
- Experience with the development or validation of LLM-as-a-judge systems
- Prior work supporting customer-facing products at scale