Data Engineer, Data & AI

Paul Street is a partner marketing agency. We run partner programs for a growing list of brands. Our operations team builds the data infrastructure, automations, and AI workflows that let account teams do their best work, and Claude is the engine we're building it all on. This role owns the layer that connects clean, reliable data to Claude, and turns that connection into skills and agents that do real work on the team's behalf. What you'll own The foundation is built. Pipelines, warehouse, and tooling are in place. Your job is to manage, scale, and extend it, then build the AI layer on top.

• Maintain, scale, and monitor the existing data pipelines. Connect new data sources as the program grows.
• Keep partner performance data clean, reliable, and documented across Impact, CJ, Partnerize, Everflow, and Awin.
• Own the context layer: feed clean data and the right context into Claude (our primary AI tool) and build the skills, projects, and automations that turn it into insight and action.
• Drive Cowork adoption across the team so people actually use what you build.
• Build agents. The goal is for Claude to do the work: many agents running proactively inside our existing workflows, handling tasks on behalf of account managers and ops, not waiting to be asked. You'll partner with us to design and ship those agents and the skills behind them.

Stack: DLT, dbt, Dagster, BigQuery, Google Cloud, with Claude / Cowork as the AI layer. What success looks like

First 30 days: own the foundation.

• You understand the full data infrastructure end-to-end; nothing is a black box.
• Pipelines are monitored with alerting, so you (and we) know something broke before a brand does.
• You've shipped one small skill or automation a team member is already using.

By 60 days: extend it.

• At least one new data source connected end-to-end and flowing into Claude.
• First proactive agent live in a real workflow, running on its own and saving a specific person measurable time.
• You've documented how to build and ship a skill so the pattern is repeatable, not locked in your head.

By 90 days: scale it.

• Multiple agents running proactively across team workflows, with adoption you can point to.
• A clear, prioritized backlog of the next agents and data sources, scoped with us.
• The team treats Claude as a teammate that does work, because you built the pipes and the agents that make it real. What we're looking for

AI & agent-building: this is the heart of the role

• You build agents. Real, working agents that do tasks proactively, not chatbots or one-off prompts. This is the single most important skill we're hiring for.
• AI is your primary coding assistant. You use Claude, Claude Code, Codex, or similar to develop infrastructure, pipelines, skills, and agents. The bar isn't how fast you type; it's how well you direct, review, and ship with AI doing much of the coding.
• You understand context engineering: getting the right data and context to Claude so what you build is accurate, reliable, and genuinely useful.
• You come in already fluent in this space; we won't be teaching the basics. You know how AI agents work and understand ML at a basic level.
• You've built automations and integrations, not just used AI for writing.
• Bonus: you've built skills, custom agents, or agentic workflows that other people actually used.

Technical

• Strong SQL and Python: strong enough to architect data systems and to read, debug, and verify code, including what your AI assistants generate. You won't hand-write everything, but you own whether it's right.
• Hands-on experience with dbt, DLT, Dagster, and BigQuery in a Google Cloud environment.
• Experience working with REST/GraphQL APIs and data modeling.
• You've built and maintained pipelines before. You're not learning the basics here, you're running them.
• JavaScript is a nice to have.

Product mindset

• Scopes before building: asks what would make something not worth building before giving a time estimate.
• Turns vague requests into clear tickets with a real definition of done.
• Flags scope changes early, not after.

Marketing domain

• Comfortable with performance metrics: CAC, LTV, CVR, AOV, ROAS.
• Knows how these metrics apply to partner programs.
• Prior experience working in or close to a marketing team is a strong plus. Who you are
• Curious: you follow the AI space because you want to, and you ask questions before making assumptions.
• Systems thinker: when someone describes a problem, you're already thinking about the full picture, not just the one fix.
• High integrity: you say what's true, own your mistakes, and don't oversell work that's still in progress.
• Accountable: if something slips, you say so before anyone has to ask.
• Fast learner who adapts quickly: you pick things up and move, you don't wait to be told twice.
• Startup experience is a plus: we move fast, and you should be comfortable in that kind of environment. How we work Ship, document, retrospect, move on. Status is visible. Scope is set before work starts. We value people who speak up about what's right and what's wrong from where they sit.