StudioPulse — The €100 Lakehouse

Multi-tenant analytics platform for fitness businesses — pilot tenant: my own two studios. A production lakehouse under a €100/month ceiling: dlt → S3 → Databricks (Delta, medallion) → dbt (Kimball) → Postgres row-level security → Superset, orchestrated by Airflow 3; Terraform + Databricks Asset Bundles; GDPR by design (Frankfurt, minimization in the extraction SELECT).

Tenant isolation as a structural property (tenant_id in every key + two RLS locks, asserted by tests). 4,277 Python + 892 dbt tests + Playwright end-to-end through real SSO.

Built AI-first: Claude Code agents directed by a decision log, quality gates and adversarial verification — 88 of 91 commits agent-co-authored, none merged without evidence. First commit → production dashboard with SSO in ~10 days.

Full write-up: the 12-post series “The €100 Lakehouse” starts on this blog on 26 August.