Member of Data Staff, Perplexity

• We're looking for someone who's been a great data scientist, analytics engineer, or data engineer: the kind of person who knows which metric actually matters, can design an A/B test that answers the real question, has gone deep on a data model because something didn't add up, and has decided that the highest-leverage thing they can do next is build AI systems that fundamentally change how data science gets done
• 6+ years in data science, analytics engineering, data engineering, or a related role. You've been close enough to real data work to know what should and should not be automated
• Deep SQL and analytics judgment - you can reason through metrics, experiments, data models, and messy warehouse reality without relying on a tool to think for you
• Strong product sense - you understand what stakeholders actually need, what makes a workflow adoptable, and how to turn a prototype into a product people use
• Production-oriented Python ability - you can build and ship working tools, wrangle APIs, evaluate model outputs, deploy services, and write code others can maintain
• Hands-on LLM experience - you've built with frontier models, agents, RAG systems, evals, or AI-powered workflows and have opinions about where they work and where they fail
• Pipeline and modeling fluency - you've worked with dbt, warehouse schemas, data quality issues, and the practical tradeoffs behind durable data systems
• Builder mentality - you see a manual process and immediately think about how to systematize it. You ship fast, measure quality, and iterate
• Autonomy - this is a new function. You'll help define the roadmap as much as execute it

Desirable

• You'll build AI agents and internal systems that can increasingly handle end-to-end analysis workflows: forming hypotheses, writing and running queries, interpreting results, and drafting recommendations with the right evaluation, review, and guardrails
• You'll build the retrieval infrastructure and evaluation loops that let AI systems query the warehouse reliably
• You'll create workflows that detect, diagnose, and help fix data issues before they become company-wide problems
• You'll build the infrastructure that multiplies what a small data team can ship
• You'll join a data team that's already using AI across its work
• The next step is turning those individual workflows into scalable systems, shared tools, and an AI-native operating model that becomes a benchmark for how modern data teams work
• Build AI agents that do data science - not just SQL copilots, but systems that can safely explore data, form hypotheses, run queries, interpret results, and generate actionable recommendations with clear evaluation and human review loops
• Make AI systems query the warehouse reliably - build the retrieval infrastructure and evaluation loops that let agents use our semantic context and metadata accurately
• Accelerate the AI-native data workflow - turn the best existing AI-assisted workflows into repeatable systems, reusable tools, and patterns the whole data team can adopt
• Automate the data lifecycle - build self-healing pipelines, automated dbt model generation and validation, data quality agents, and diagnosis workflows that reduce manual firefighting
• Ship AI-powered experiment analysis - build agents that interpret A/B test results, flag statistical issues, identify likely drivers, and draft ship/no-ship recommendations
• Turn the data team into a product team - build internal data products that stakeholders use every day, replacing ad hoc requests with self-serve AI interfaces
• Own the full lifecycle - identify high-leverage problems, prototype with LLMs, evaluate accuracy, design the UX, ship to production, and monitor quality over time
• Set the standard for the industry - Perplexity's data team is already using AI across its work. You'll turn that into something other data teams look to as the benchmark
• Build AI with AI - Perplexity builds AI for people who expect more. You'll bring that same level of ambition to how the company works with data
• Frontier models, day one - you're at an AI company with access to frontier infrastructure and people who deeply understand what's possible
• Massive leverage - the systems you build will multiply the output of every data team member and every stakeholder who needs data
• Direct impact - small team, no layers of approval. Idea to shipped system in days, not quarters