Anthropic · job-boards.greenhouse.io · checked today
Security Data Engineer
Anthropic’s mission is to create reliable, interpretable, and steerable AI systems. We want AI to be safe and beneficial for our users and for society as a whole. Our team is a quickly growing group of committed researchers, engineers, policy experts, and business leaders working together to build beneficial AI systems.
Skills, with evidence
- Data modelling
Have integrated with systems you did not control (inventory tools, ticketing systems, vendor APIs) and built data models that outlive any one source
must have - Data quality
Data quality: build coverage checks, reconciliation against sources and alerts on a silent or drifting feed, so a wrong or missing record is caught the same day; own retention and access controls on the records
must have - SQL
Write SQL and at least one general-purpose programming language well
must have - Streaming
Have designed and run production data systems (relational schemas, change data capture, batch or streaming pipelines and the services around them) and kept them correct
must have - Airflow / orchestration
build the pipelines, stores and tools, and run what you ship
- Failure handling
Security at that scale runs on data, and you'll be the engineer who owns it and the systems that produce it: component-level hardware inventory and chain of custody, fed by software that runs at our sites and on hosts, and continuous control monitoring, with the posture calculations that decide what each site is cleared to run.
- Governance & security
Continuous control monitoring and posture: own the pipelines that turn check outcomes, assessment results and findings into per-control and per-site measures over time, against the compliance frameworks we and our partners are assessed against, and the posture calculation that decides what a site is cleared to run
not practised here - Incremental loads
Have designed and run production data systems (relational schemas, change data capture, batch or streaming pipelines and the services around them) and kept them correct: coverage measured, records reconciled against the source, alerts on silence, and a design that expects records to get corrupted or lost
- Warehousing
Experience with asset management, supply chain or warehouse systems where serialized tracking and custody transfer are the core problem
Your plan
- The SQL screen: correct, then fast≈ 1 h
Data quality · SQL
- Average order value by countryFoundations
- Line items per orderFoundations
- Revenue by buyer countryFoundations
- Customers who never orderedFoundations
- Hiring funnel by roleFoundations
- Data modelling: the round most people fail≈ 2 h
Data modelling · Warehousing
- Marketplace core entitiesFoundations
- Cinema seat bookingFoundations
- City parking baysFoundations
- Dating app matchesFoundations
- Food delivery ordersFoundations
- Pipeline design: safe to run twice≈ 3 h
Streaming · Airflow / orchestration · Failure handling · Incremental loads
- After the first one finishesFoundations
- Run it for last TuesdayFoundations
- The history nobody keptFoundations
- The spreadsheet is a dependencyFoundations
- Too slow by morningFoundations
- Say it out loud≈ 1 h
15 drills · Foundations + Intermediate≈ 6 hours
Not covered by the plan: Governance & security.
Readiness
Counted from drills you have completed anywhere on D8LooP.
leaves in 2 dremoved the moment Anthropic closes it