Platform engineer based in Rio de Janeiro. I design and operate Kubernetes platforms for teams that ship to production every day, mostly in financial services.
Over the last years that meant running a fleet of 330 clusters behind ArgoCD, Atlantis and Rancher, turning EKS provisioning from a two-day manual process into a 40-minute Terraform/Terragrunt pipeline, moving a 3,000 jobs/day Jenkins farm to immutable infrastructure, and treating cloud cost as an engineering metric rather than a finance report.
I care about boring, predictable infrastructure: everything in code, every change reviewed, upgrades that nobody notices.
Open source
Contributing upstream across the cloud native ecosystem: CNCF projects like Kubernetes, Envoy, Prometheus, Argo, Crossplane, Headlamp and CoreDNS, plus Terraform providers, Helm charts and AI infrastructure tooling (NVIDIA device plugin, Ray, deepwiki-open).
Merged
- prometheus-community/helm-charts: [prometheus-pgbouncer-exporter] Fix ServiceAccount annotations and NetworkPolicy labels rendering
- prometheus-community/helm-charts: [prometheus-pushgateway] Fix accessModes indentation in StatefulSet volumeClaimTemplates
- prometheus-community/helm-charts: [prometheus-postgres-exporter] fix indentation of serviceAccount annotations and networkPolicy labels
- prometheus-community/helm-charts: [prometheus-nginx-exporter] fix ServiceMonitor rendering of labels, targetLabels and relabelings
- prometheus-community/helm-charts: [prometheus-pgbouncer-exporter] Pass config.extraArgs to the exporter
- fluent/fluent-operator: fix(helm): honor logstashFormat: false in the Elasticsearch ClusterOutput
- openobserve/terraform-provider-openobserve: fix: keep the configured JSON text when the server document matches
- prometheus-community/helm-charts: [prometheus-blackbox-exporter] Set port names on self ServiceMonitor endpoints
- prometheus-community/helm-charts: [prometheus-operator-admission-webhook] Fix serviceMonitor.metricRelabelings value name
- rook/rook: osd: don't wipe metadata devices of existing OSDs
- aminueza/terraform-provider-minio: fix(replication): reject two rules on the same target endpoint
- kubernetes-sigs/dra-example-driver: Reject GPU sharing configs that omit the strategy config instead of panicking
- lima-vm/lima: limactl edit: warn when changing the VM type of an existing instance
- lima-vm/lima: limayaml: don't panic when the latest config fails to load
- sergelogvinov/proxmox-cloud-controller-manager: fix: remove the right addresses when several are ignored
- fluent/fluent-plugin-kafka: out_kafka2: Unsubscribe ActiveSupport notifications on shutdown
- sergelogvinov/proxmox-cloud-controller-manager: fix: keep ip_sort_order cidrs out of the ignore list
- kubernetes-sigs/headlamp: backend: auth: Accept a single string groups claim in /me
- tsuru/charts: fix(tsuru-prometheus-pool): honor securityContext from values
- tsuru/charts: ci: publish charts as OCI artifacts to GHCR
- openebs/lvm-localpv: feat(chart): allow extra args for the controller CSI sidecars
- fluent/fluent-plugin-kafka: in_kafka_group, in_rdkafka_group: Fix records skipped with time_source record
- hetznercloud/hcloud-cloud-controller-manager: fix(load-balancer): trim spaces in comma separated annotations
- openobserve/openobserve-helm-chart: fix(openobserve-collector): apply image.pullPolicy to the collectors
- AsyncFuncAI/deepwiki-open: feat: add Helm chart
- kubernetes-sigs/headlamp: frontend: resourceMap: Resolve HTTPRoute parent and backend refs
- kubernetes-sigs/headlamp: frontend: util: Use desired DaemonSet pods as the total
- tsuru/tsuru-client: fix: keep the current target in sync on target update
- coredns/helm: docs: note that the ServiceMonitor needs the metrics Service
- kubernetes-sigs/headlamp: frontend: resourceMap: Match NetworkPolicy podSelector fully
In review
Projects
- helm-charts: production-ready Helm charts with secure defaults, Gateway API, HPA/KEDA and schema-validated values, tested and released from CI
- terraform-aws-platform: opinionated Terraform modules for EKS platforms (VPC, EKS with access entries and Pod Identity, Karpenter), tested with mock providers and laid out for Terragrunt
Certifications
