Analytics Engineer

About the role

We're building a modern analytics platform from the ground up: a layered, well-tested data model in Snowflake that becomes the single source of truth for the business, and a governed semantic layer that lets stakeholders, not just analysts, ask questions and get trustworthy answers. As an Analytics Engineer, you own the transformation layer. You'll turn messy source data from a dozen operational systems into canonical facts and dimensions, migrate legacy reports onto that clean foundation, and make the whole thing reliable enough that people stake real decisions on it. How we work matters as much as what we build. We leverage AI heavily across our development workflow, authoring and refactoring SQL, building dbt models, writing tests, and moving work through our GitHub PR process. AI is a force multiplier here, not a crutch and not a black box. You own everything that ships under your name. That means you read every line, understand why it's correct, catch what the model got wrong, and stand behind the result in review. We don't write everything by hand anymore, but you can't own what you don't understand, so strong fundamentals are non-negotiable.

Responsibilities

• Build the Core data model: design and implement facts and dimensions in dbt following a disciplined staging → intermediate → core → mart architecture; model slowly changing dimensions, handle mixed-grain snapshot sources, and make defensible grain and materialization decisions.
• Migrate legacy reporting onto Core: reconstruct existing business-critical views and reports on top of the new model, reconciling outputs line-for-line so stakeholders can trust the cutover.
• Work fluently with AI in the loop: use AI coding tools to accelerate model development, test writing, and the dbt/GitHub workflow, while critically reviewing every output, correcting it, and taking full ownership of correctness, performance, and style.
• Own data quality: write dbt tests (generic, singular, and unit), establish contracts on data-out models, and treat a failing CI check as a release blocker, not a suggestion.
• Integrate diverse sources: work across data from a myriad of sources, each with its own quirks, grains, and coverage gaps you'll need to understand and document.
• Build the semantic / AI-ready layer: curate mart models and metric definitions with the metadata (certification, PII level, known issues) that powers governed self-service and agentic AI, so non-analysts can safely ask their own questions.
• Raise the bar on engineering practice: small, reviewable PRs; a shared style guide; version control as the