Principal Site Reliability Engineer | Cloud & Platform Engineering
I build the systems and engineering practices that make production easier to operate: reliable services, shared cloud platforms, safe delivery and faster incident response.
LinkedIn · Portfolio · Featured project: AI SRE RCA · Português
I'm a Principal SRE and former SRE Tech Lead with 13+ years in infrastructure, distributed systems and platform engineering. I've worked hands-on across AWS, Azure and GCP, while setting technical direction and engineering standards for teams operating critical services. My background ranges from 24×7 hosting to global developer platforms and a regulated fintech.
| Scope | Documented impact |
|---|---|
| Platform adoption - Ambev / AB InBev | Backstage-based internal platform used by 250+ squads, with 800+ services cataloged across Latin America and Europe |
| Developer experience - Ambev / AB InBev | New-service onboarding reduced from ~2 weeks to under 1 day through reusable templates and self-service |
| Incident recovery - Caju | MTTR reduced from ~120 to ~40 minutes (-67%); critical-service availability improved from 99.5% to 99.9% |
| Delivery - Caju | Change lead time reduced from ~2 weeks to under 2 days through GitOps and progressive releases |
| Cloud efficiency - Caju | AWS costs reduced ~28% (~BRL 45k/month) through rightsizing, autoscaling and resource cleanup |
These results belong to the roles described below, not to my personal GitHub projects.
- Led the development and adoption of a Backstage internal developer platform with Golden Paths, service catalog, documentation and reusable self-service templates; worked across engineering organizations in Latin America and Europe.
- Designed and operated multi-cloud Kubernetes platform capabilities across Azure AKS, AWS EKS and GCP GKE. Standardized Infrastructure as Code and CI/CD with Terraform / OpenTofu, Azure DevOps and GitOps.
- Partnered with security, enterprise architecture and FinOps stakeholders to turn requirements into platform guardrails and 20+ adopted RFCs/ADRs. Mentored engineers and technical leaders across regions.
- Worked across 12 squads, ~60 microservices and five EKS clusters in a regulated fintech, combining hands-on reliability work with technical leadership.
- Established SLIs/SLOs, error budgets, actionable observability, runbooks, postmortems and on-call practices. Improved incident detection from ~30 to ~2 minutes and reduced recurring incidents by ~40%.
- Improved release safety with Argo CD, Flux, Argo Rollouts and Flagger; expanded developer self-service with Terraform-based infrastructure patterns.
- Worked with PCI-DSS requirements, identity/access controls, secrets and security guardrails in production operations. This does not imply ownership of a PCI audit or certification.
- Helped modernize high-traffic, 24×7 hosting systems; led migration of ~80% of production workloads from on-premises infrastructure to AWS.
- Automated provisioning and standardized Linux infrastructure, improving the operational consistency of a fleet of 1,000+ servers.
Featured engineering work: AI SRE RCA
A controlled, evidence-first incident investigation system for Kubernetes and distributed applications. I built it to explore a practical engineering question: how do we use AI to accelerate investigation without letting an LLM invent the root cause or cross security boundaries?
| Design decision | Implementation |
|---|---|
| Keep diagnosis verifiable | Deterministic RCA links confirmed causes to integrity-checked Kubernetes/telemetry evidence; missing evidence yields an inconclusive outcome |
| Use AI as an assistant, not an authority | Optional Copilot explains verified findings and cites evidence; model output cannot change persisted verdicts or trigger remediation |
| Constrain the investigation | Authorized tenant/cluster/workload and time scope, read-only collection, query budgets, persisted jobs, replayable SSE progress |
| Separate action from analysis | Remediation lives behind its own permissions, policy checks and explicit human approval |
| Test failure modes | Documented synthetic benchmarks and controlled Kind scenarios, including multiple independent causes and correct abstentions |
Screenshot: controlled local Kind investigation displayed as a read-only result; not a production incident. The project is a development release candidate, not a production deployment.
Architecture and security boundaries · Running UI screenshots · Reproducible validation reports
- Production Readiness Framework: service reviews for SLOs, overload behavior, failure modes and recovery.
- Platform Engineering Reference: Backstage, Kubernetes and Argo CD patterns for developer self-service.
- Kubernetes Incident Evidence Collector: read-only Python collector; mock-tested prototype, live-cluster validation pending.
- Observability Stack: local OpenTelemetry, Prometheus, Loki, Tempo and Grafana environment.
- Terraform Cloud Platform: modular Terraform/OpenTofu infrastructure examples.
These repositories are reference implementations, prototypes and labs; each has its own documented limitations.
- Reliability is operational: define what matters through SLIs/SLOs, make changes observable, and rehearse recovery before incidents.
- A platform succeeds when teams adopt it: keep the default path safe, documented and easier than bespoke infrastructure.
- Security is part of the design: enforce isolation, least privilege, reviewable changes and explicit operational boundaries.
- Evidence beats confidence: distinguish observed facts from hypotheses; an inconclusive diagnosis is better than an invented one.
Core stack: AWS · Azure / AKS · GCP · Kubernetes · Terraform / OpenTofu / Bicep · Backstage · Azure DevOps · Argo CD · Python / Go · OpenTelemetry · Prometheus / Grafana · Datadog · PCI-DSS-aware operations.
Credentials earned: CKA (2023) · AWS Solutions Architect - Associate (2023) · Istio Certified Associate (2022) · Terraform Associate (2022). Current renewal/expiration status is not asserted.
Languages: Portuguese (native) · English (advanced professional) · Spanish (limited working proficiency).




