Event pipeline
Kafka-based ingestion with schema registry and replay support.
Microservices · Kubernetes · Real-time dashboards
SaaS Analytics Platform
A B2B analytics product ingesting millions of customer events daily, with per-tenant isolation, sub-second dashboards, and deploys the team finally trusts.
Challenge
One noisy tenant could slow everyone down.
The original single-database design meant a single customer's event spike degraded dashboards for all tenants. Releases were monthly, manual, and feared — the team had stopped shipping on Fridays entirely.
Approach
Isolate tenants, automate everything.
We split ingestion, processing, and querying into dedicated services with tenant-aware routing and per-tenant rate limits. A GitLab pipeline with parallel test stages and canary deploys to Kubernetes took releases from monthly to multiple per week.
Deliverables
Tenant isolation
Routing, quotas, and noisy-neighbor protection per customer.
CI/CD & GitOps
Canary releases with automated analysis and one-click rollback.
Observability
Distributed tracing and per-tenant dashboards from day one.
Weekly
releases, up from monthly
Minutes
to roll back a bad deploy
Isolated
tenants — spikes stay contained