Perplexity

This role focuses on building and maintaining automated evaluation pipelines to assess the quality of answers generated by LLMs in Perplexity's search engine and other products. It involves designing evaluation sets, developing VLM-based solutions for visual rendering assessment, and adapting public benchmarks. The goal is to directly shape product changes by providing key evaluation metrics.

Software Engineer, Data Flywheel role focused on building systems and pipelines for accessing and utilizing eval verdicts, turning raw signals into datasets, and creating a simulator pipeline for user interactions. This role emphasizes data trust through monitoring and quality checks, operating within a high-impact team to improve Answer Quality.