Member of Technical Staff (Search Quality Analyst)
Where this listing came from
This job was taken from a public posting. Perplexity has not claimed the listing yet, so nobody there has committed to interviewing through VoxJobs yet — we interview you, we reach out to the company with the candidates, and you are told the moment they join.
About the role
Perplexity is looking for an experienced analyst to help us build and improve our core search technologies. You'll work at the intersection of data analysis and engineering - designing metrics, building data pipelines, and improving the quality of our search and answer systems.
This role is hybrid in Belgrade, London or Berlin.
Responsibilities
- Find and diagnose quality issues in our search pipeline
- Design metrics from scratch to track and measure search quality
- Build datasets for model training, including LLM-as-a-judge labeling pipelines
- Improve search snippet quality and page selection algorithms for indexing
- Design and analyze A/B experiments to validate improvements
Qualifications
- 4+ years of experience as a data analyst or in a related role
- Strong coding skills — expected to write production-grade code at a mid-level backend engineer level
- Proficiency with SQL and Python
- Demonstrated hands-on experience with at least one of the following:
- Designing metrics from scratch (not just analyzing existing A/B experiments)
- Building labeling pipelines using LLM-as-a-judge
- Training ML models that shipped to production with measurable metric improvements
- Designing evals with known ground truth (e.g. SimpleQA, BrowseComp) or driving meaningful improvements on such evals
Nice to have
- Experience working on search-related products
- Knowledge of statistics and A/B experiment design
- Experience with Apache Spark or Databricks
How you will be assessed
Your CV and written answers against the rubric of this employer. The weights are theirs, and they are published right below.
The video conversation: how you think and explain. The AI assesses what you say — the video exists for a person to watch, and is never analysed by any model.
Every sub-score cites the transcript excerpt or CV line that produced it. You see your band and the reasoning — always.
Earns points
Adds to your score. Never the reason for a rejection.
- Machine learning · weight 13%
- LLMs and generative AI · weight 13%
- Data engineering · weight 13%
- Statistics · weight 13%
- Apache Spark · weight 13%
About this company
Profile assembled from public sources — it was not written by the company, and they have not confirmed anything here yet.