Work that
held up.

Registrar-level block on the primary domain in one territory — not an application-layer failure. Standard infra rollback paths were irrelevant; the application was healthy, the name resolution was not.

We provisioned secondary routing via CloudFront without touching the application layer, switched edge resolution to a backup domain, and coordinated the DNS propagation window to minimize player-facing disruption. In parallel, we began designing a domain-agnostic architecture so a single registrar event could not take the platform offline again.

• Full platform availability restored before kickoff.
• Zero revenue loss during the event window.
• Domain-agnostic architecture delivered within 3 weeks post-incident.

Edge routing decoupled from any single registrar; multi-domain failover wired into CloudFront with automated health checks and a documented 5-minute switchover runbook.

14 manual deployment steps with no CI/CD, no environment templating, and no self-service path for product engineers. Every operator tenant required hand-rolled provisioning, so headcount — not architecture — was the throughput ceiling.

We introduced a CI/CD framework using GitLab pipelines and Helm charts, standardized environment provisioning across all operator tenants, and built operator-scoped deployment runbooks so product engineers could trigger controlled releases without DevOps involvement. The original engineer moved from operator to reviewer.

• Deployment lead time reduced from 3 days to 4 hours.
• Single-engineer dependency eliminated across all 50 operator environments.
• The original DevOps engineer moved into a platform architecture role.

Tenant provisioning fully templated in Helm; product engineers self-serve deploys behind a review gate; on-call no longer blocks shipping.

Stack assumed a single offshore region with no Brazilian data residency, no multi-region failover, and no load profile for football-event traffic peaks. SIGAP technical requirements could not be met by patching — the data and routing layers needed rebuilding.

We ran a 5-day infrastructure audit and produced a gap analysis against SIGAP technical requirements. We rebuilt the data layer with Brazilian data residency compliance, deployed multi-region failover with AWS São Paulo as primary, introduced geo-routing at the CDN layer, and ran a full load test simulating peak betting volume for a first-division football match. Documented runbooks and on-call procedures handed over on day 38.

• Go-live on day 42. Fully compliant SIGAP stack.
• Zero critical incidents in the first 90 days of operation.
• Platform held during two major football event traffic peaks.

Brazilian primary with documented failover region, residency-compliant data layer, and load-tested capacity envelope sized to peak match traffic.

23% of spend tied to three categories: orphaned resources from a migration 18 months prior, Datadog misconfigured on the startup plan with hourly billing anomalies, and RDS instances over-provisioned at under 20% utilization. No tagging discipline meant cost ownership was invisible.

We ran a full cloud cost audit across all AWS accounts and services, cleaned the Datadog configuration, rightsized the RDS layer, and decommissioned 14 unused infrastructure components. Every change was documented with rollback procedures, and a cost-visibility dashboard was wired into the CTO's weekly review.

• €40k annual savings identified and implemented.
• Datadog bill reduced by 67% without losing observability coverage.
• CTO had full cost visibility dashboard within 2 weeks.

Tagged resources, monthly cost-review cadence, and a rightsized database tier. Spend now tracks utilization, not historical accident.

Single-person knowledge dependency. No documented architecture, no on-call rotation, no escalation paths, and fragile components running on self-managed services that required operator-specific tribal knowledge.

We embedded with the team for the first two weeks, ran a structured knowledge extraction with the departing CTO, documented the full infrastructure in a runbook, introduced a basic on-call rotation with defined escalation paths, migrated the most fragile components to managed services, and ran two tabletop incident exercises. We stayed on as the infrastructure owner for the following four months while the company hired a new technical lead.

• Zero platform incidents during the transition period.
• Four operator SLA commitments maintained without interruption.
• New CTO onboarded to a fully documented infrastructure in week one.

Documented runbooks, on-call rotation in place, fragile components migrated to managed services, and a clean handover package for the incoming CTO.