AI and model teams

The Archive Keeper

For a research lead whose results must still be readable next year.

One signal · 7 specialists · one result
Research Data ArchivistRun Verification ClerkStored Data Integrity CheckerData Version PromoterStorage Incident HandlerRun Conditions RecorderStorage Spend Reconciler
  • Keeps knowledge2
  • Runs infrastructure3
  • Audits and guards2
How the work moves

Five steps, each done by one specialist

Starts with

Data left behind by a large run

First
1Checks the finished run against the original requestRun Verification Clerk2Re-checks stored files for damageStored Data Integrity Checker3Switches readers to a new version of the dataData Version Promoter
Then
4Keeps the previous version until the new one is confirmedResearch Data Archivist5Reports what a storage failure affected and what was recoveredStorage Incident Handler
Ends with

Complete, undamaged results still readable next year

the signalthe resultEach step is done by one specialist on the crew

The story

Every large run leaves data behind, and the agent looks after it. It checks a finished run against the original request and reports whether every item is present. Stored files are re-checked for damage. When a new version of the data is built, the agent switches readers to it and keeps the previous one until the new one has been confirmed. If storage fails, it reports what was affected and what was recovered.

The crew

The specialists behind this role, grouped by the kind of work they do.

Runs infrastructure
Keeps knowledge
Audits and guards

Where this crew also works

Each bar counts the specialists this role shares with another.

The engineering behind this role

GPU vector search and corpus analyticsCapabilityHigh-throughput batch inference and embeddingsCapabilityAppend-only, with tombstones and lineageIdeaThree shards that move togetherEngineering storyMillions of queued items that were not thereEngineering storyAll capabilitiesThe engineering, by what it does