Staff Studio
Staff and UpThe rounds that decide Staff offers: a deep dive on a project you led, a critique of someone else's design, and judgment calls without a clean answer.
Design Docs
Design Doc · Platform Team
Move Notifications to a Shared Event Bus
Fourteen per-team queues become one Kafka topic and a central dispatcher for email, push, and SMS. Review the delivery guarantees, the rollout, and the isolation story.
7 Sections · 25 Min
Review This DocDesign Doc · Catalog Team
Add a Read-Through Cache to the Catalog API
A Redis cache in front of Postgres for product reads. Review the cache key, the invalidation order, and what happens when Redis is slow or gone.
7 Sections · 25 Min
Review This DocDesign Doc · Identity Team
Move User Data Into Its Own Service and Database
The users, profiles, and credentials tables move out of the monolith's shared Postgres into a User Service with its own database. Review the migration plan, the foreign keys other tables hold, and the cost of the new network hop.
8 Sections · 25 Min
Review This DocDesign Doc · Checkout Team
Run Checkout Active-Active Across Two Regions
Checkout moves from one region with a cold standby to active-active in us-east-1 and eu-west-1. Review the conflict resolution, the failover plan, and the availability target.
8 Sections · 25 Min
Review This DocDesign Doc · Billing Team
Replace Nightly Batch Billing With Streaming Usage Metering
Usage events flow through Kafka and Flink into a live meter that replaces the nightly batch rating job. Review the event correctness, the reconciliation plan, and how customers see the change.
7 Sections · 25 Min
Review This DocDesign Doc · Search Team
Index Product Search From Change Data Capture
Product search moves from a nightly full reindex to a Debezium change stream from the catalog Postgres database. Review the recovery path, the event ordering, and the load on the primary.
8 Sections · 25 Min
Review This DocDesign Doc · Auth Platform Team
Replace Server Sessions With Stateless Tokens
Session lookups in a shared Redis store give way to signed JWTs that each service verifies locally. Review the signing scheme, revocation, and how the cutover treats signed-in users.
6 Sections · 25 Min
Review This DocDesign Doc · Infrastructure Team
Cut Compute Spend by 30% Before Q3
Spot instances, rightsizing, a single availability zone, and three year commitments to cut compute spend by nearly a third. Review what each change risks and whether the savings add up.
5 Sections · 25 Min
Review This DocDesign Doc · Developer Platform Team
Consolidate Three Deploy Systems Into One Platform
Jenkins, Spinnaker, and the data platform's scripts give way to one platform built on Argo CD. Review the deadline, what the platform covers at launch, and how teams roll back.
4 Sections · 25 Min
Review This DocDesign Doc · Feed Ranking Team
Launch a Ranking Model for the Home Feed
A click prediction model replaces recency ranking on the home feed. Review the objective, the features and training data, the launch plan, and what happens when the model is slow.
5 Sections · 25 Min
Review This DocDesign Doc · SRE Team
Adopt SLOs and an Error Budget for the Public API
Threshold alerts give way to service level objectives, SLO paging, and a deploy freeze when the budget runs out. Review the indicators, the targets, the alerting, and the policy.
6 Sections · 25 Min
Review This DocDesign Doc · Orders Team
Shard the Orders Database
One 6 TB Postgres primary splits into 12 shards. Review the shard key, order IDs, cross-shard transactions, the migration window, and how the design grows.
6 Sections · 25 Min
Review This Doc