Member of Technical Staff (Data Scientist, Evals)

Perplexity; United States; Remote; USD 100,000 - 180,000; Full time

Job summary

Perplexity seeks a Data Scientist to enhance answer quality by building and maintaining automated evaluation pipelines for its LLM-first search engine. This fully remote role focuses on creating robust evaluation protocols that drive product improvements and user satisfaction.

You will work with a small, high-impact team to adapt public benchmarks for practical product performance, design evaluation sets for tool usage, and ensure rendering quality across platforms.

Qualifications

• Strong background in data science
• Proficiency in Python and SQL
• Experience designing and running evaluation experiments

Responsibilities

• Architect and maintain automated evaluation pipelines to assess answer quality across Perplexity's products
• Develop evaluation sets to measure the impact of tool calls on answer quality
• Evaluate visual rendering across platforms and ensure cross‑platform consistency
• Collaborate with technical leadership to adapt benchmarks and evaluations for practical product performance measurements

Skills

Python

SQL

Data science

Job description

Perplexity seeks a Data Scientist to enhance answer quality through specialized evaluations in a fully remote role. Candidates with a strong data science background and proficiency in Python and SQL should apply.

In this role, you will architect and maintain automated evaluation pipelines that assess the quality of answers across Perplexity's products. Your work will directly influence the accuracy and helpfulness of responses generated by their LLM-first search engine, ensuring that the evaluations align with user needs.

You will design evaluation sets to measure the impact of tool calls on answer quality and develop solutions to evaluate visual rendering across platforms. Working within a small, high-impact team, you will collaborate with technical leadership to adapt public benchmarks and academic evaluations for practical application in product performance measurements.