SRE/Cloud Engineer
EvBox | Amsterdam
- Built a Ray-orchestrated, multi-node vLLM inference cluster as a proof of concept (DeepSeek-R1): distributed serving across smaller GPUs, where calls waited for the cluster to scale up before inference began.
- Owned the Prometheus and VictoriaMetrics observability platform: a central, multi-tenant metrics cluster with HA storage and auto-scaling ingestion, serving as the metrics backbone for 30 microservices.
- Built the core GitOps platform managing the full lifecycle for 30 microservices across 8+ GKE clusters and 1,000+ nodes, with IAM, DNS, and Vault authentication bootstrapped from scratch in Terraform.
- Built the company’s HashiCorp Vault deployment from scratch: PKI-based mTLS and KV v2 handling 500+ requests per second, eliminating static credentials across three environments.
- Designed the Vault-backed custom PKI and Pomerium identity-aware access layer that replaced Google BeyondCorp: certificate-based client validation for every cluster, without its licensing cost.
- Made SBOM creation, scanning, and lifecycle management a first-class stage of the container image pipeline.
- Productized CI for 30 product teams: shared GitLab CI/CD pipeline templates and 24 custom base images enforcing Snyk and SonarQube quality gates.
- Built ephemeral per-branch preview environments and a two-phase GCP project factory with least-privilege custom roles and automatic budget guardrails.
- Operated production on-call as incident commander in a 5-person SRE team embedded with product teams across the company, on Google’s SRE methodology, and mentored junior engineers along the way.