Work · Operations
Atlas
One console for service health, deploys and access control, replacing eleven spreadsheets and three internal tools.
Client
Confidential · infrastructure team
Sector
Operations
Duration
24 weeks
Our role
Product design, frontend, platform engineering
atlas / services
Service health
Uptime
99.98%
+0.02
p95 latency
142 ms
−18 ms
Error rate
0.031%
−0.004
14:20 UTC
1,284 req/s
| Service | Env | Region | p95 | Status |
|---|---|---|---|---|
| ingest-api | production | eu-west-1 | 142 ms | ok |
| ledger-sync | production | us-east-1 | 88 ms | ok |
| report-builder | staging | eu-west-1 | 310 ms | degraded |
| auth-edge | production | global | 24 ms | ok |
| billing-jobs | production | eu-west-1 | 1.2 s | queued |
11 → 1
spreadsheets retired
−72%
time to answer an access question
2 days
engineer onboarding, down from 2 weeks
100%
privileged actions audited
The challenge
Every team kept its own spreadsheet. Nobody could answer basic questions — who has access to production, what shipped last night, which service is degraded — without asking three people. Onboarding an engineer took a fortnight.
Our approach
We interviewed twelve engineers, mapped the eleven spreadsheets to five real objects, and designed one console around those objects. We shipped weekly behind flags, migrating one spreadsheet at a time so no team lost their workflow overnight.
Deliverables
What we shipped
- 01Unified operations console with service, deploy and access views
- 02Role-based access control with full audit trail
- 03Deploy timeline with rollback from the interface
- 04Migration tooling for eleven legacy spreadsheets
- 05Design system and component library handed to the in-house team
Technology
- React
- TypeScript
- Go
- Postgres
- AWS
- Terraform