About the Role

Moe's price is only as good as the comps behind it, and real-world sold-listing data is messy by default — duplicate listings, wrong categories, faked or manipulated prices, inconsistent condition grading. You'll build the pipeline that turns that mess into comps a user (and an investor) can trust. This is arguably the least visible and most decision-critical part of the product: get it wrong and every price Moe shows is quietly untrustworthy, no matter how good the identification model is.

What You'll Do

• Build and own ingestion pipelines pulling sold-listing data from multiple marketplaces and sources, at scale and on a schedule
• Design entity resolution logic that matches messy, inconsistent listing data (text + images) back to a canonical item
• Build the valuation logic that turns a set of matched comps into a price estimate — weighting recency, condition, and comp count appropriately
• Build data quality monitoring that catches drift, fraud, and stale comps before they reach a user
• Partner with the CV lead on what "the same item" means across data sources with different levels of detail
• Make build-vs-buy calls on data sources, licensing deals, and third-party pricing feeds

What We're Looking For

• 5+ years in data engineering, with direct experience building and owning large-scale ETL/data pipelines in production
• Real experience with messy, adversarial, real-world data — marketplace, pricing, fraud, or a similarly noisy domain
• Strong SQL and a modern data stack (dbt, Airflow or Dagster, Spark or similar)
• Some exposure to statistical estimation or pricing models — even basic regression-based comps logic counts
• Comfortable owning ambiguous data quality tradeoffs with no perfect answer
• Strong communication — you'll need to explain confidence and uncertainty in pricing to non-technical stakeholders

Nice to Have

• Background in pricing intelligence — real estate AVMs (Zillow Zestimate), used-car pricing (Carvana, CarGurus), or resale marketplaces
• Experience with entity resolution / record linkage at scale
• Experience negotiating or managing third-party data licensing relationships
• Founding or early-stage startup experience

What Success Looks Like

• 30 days: Fully ramped on available data sources; has ingested and profiled data for the first category
• 60 days: Working comps pipeline live for one category — ingest, dedupe, match, price — with a visible confidence range
• 90 days: Pipeline extended to 2–3 categories, with data quality monitoring catching bad comps automatically

Process

Intro call → technical deep dive on a past pipeline you've built → a scoped take-home or pairing session on a real Moe data-matching problem → founder conversation.