"Is this design sound?"
Architecture review. A second opinion on a design or an existing estate before you commit to it: failure modes, security posture, operability and cost, written down with priorities.
$ whoami
I'm Fabio Gimenez Eleuterio, an AWS Certified Solutions Architect – Professional. I have spent 18 years in infrastructure, from data center floors and BGP sessions to cloud and data platforms in regulated fintech and banking. I have designed it, run it in production, and answered for the bill.
$ man engage
There is no catalog here. These are the questions people most often bring me, and the shape the help tends to take.
Architecture review. A second opinion on a design or an existing estate before you commit to it: failure modes, security posture, operability and cost, written down with priorities.
Cost clarity. Read-only analysis of your billing data: every significant movement traced to its cause, commitment coverage, waste, and a short list of decisions with a monthly number on each.
Migration and hybrid plan. Target architecture, connectivity, sequencing and cutover for a move to AWS, between regions or engines, or to a hybrid setup that has to keep the data center.
Reliability and DR check. Tiers, backups, restore tests and scaling behaviour compared with what the business actually needs, and the gaps ranked by risk.
Audit readiness. Mapping SOC 2, ISO 27001 or PCI DSS requirements onto your AWS setup, and turning findings into engineering work that fits a roadmap.
Sounding board. A recurring conversation for a CTO, a head of platform or a team lead who wants an experienced architect to think out loud with.
Not sure which one fits? Describe the problem in a few lines. I will tell you honestly whether I am the right person for it.
$ cat results.log
Company names are left out on purpose. The figures are real, taken from the billing data, pipeline metrics and audit reports of platforms under my responsibility.
FinOps · unit economics
On a fintech platform, AWS cost per dollar of revenue fell 13% in a year, and 11% counting the data platform, while business volume grew about 30% and new environments and AI workloads came online. Cloud and data together ran at about 2% of revenue. The lever was a unit-cost model that finance could trust: billing data loaded into the data platform, and cost visible per owning team.
FinOps · commitments
Compute Savings Plan coverage went from roughly 70% to 96% with utilization above 90%. A relational database fleet moved to Graviton and was re-reserved, with reservations 100% utilized. Commitments are reviewed against the roadmap, so a planned rightsizing never leaves paid capacity idle.
FinOps · waste
Non-production spend cut by 25% with scheduled scaling and smaller databases. Load balancer spend halved (−48%) by removing old gateways and rightsizing the mesh ingress. Kubernetes clusters consolidated, taking control plane cost down 72%. Legacy database engine versions upgraded, removing extended-support charges almost entirely.
FinOps · guardrails
Cost allocation tagging completed, then more than 500 budgets across services and scopes, an organization-wide anomaly monitor with daily alerts, and a monthly cost report inside the infrastructure review. When an analytics workload spiked, it was traced to its owning team and contained within weeks.
Platform · delivery
Pipeline speed set as a platform KPI and tracked on a DORA dashboard. The median went from 10.6 minutes to 6.1 in four months. On that platform, newly hired engineers shipped their first change within eight days of starting.
Reliability · databases
A major version upgrade of production PostgreSQL completed with under a minute of downtime, using blue/green deployments on Multi-AZ. A disaster recovery tier framework, with a tested snapshot restore at 26 minutes against a 40-minute target. Connection pooling that cut database connections by 98%.
Security · compliance
In regulated environments, platforms under my responsibility passed SOC 1 and SOC 2 Type 2 with no exceptions and an ISO 27001 surveillance audit, with seven controls moving from minor non-conformity to compliant. For PCI DSS: network diagrams, a segmentation test environment and a prioritized gaps register that became roadmap work.
Data platform
Change data capture from production PostgreSQL with Debezium and Kafka into S3 and the warehouse, with operational, lake and warehouse layers built with Airflow and dbt. Message loss traced to a shared Kafka client library and fixed at the root. PII masked for every downstream consumer without moving to a more expensive warehouse edition.
Hybrid · migrations
Private connectivity from on-premises environments to AWS Direct Connect, Azure ExpressRoute and Google Cloud Interconnect, designed and delivered from inside the interconnection industry. Migrations to the cloud sequenced so that the network was ready before the first workload moved, followed by cost optimization of the estates that resulted.
$ ls ./capabilities
I'm most useful where architecture, operations and money meet. These are the areas where I have done the work myself, led the teams that did it, and lived with the result afterwards.
Multi-account foundations, networking, compute on EKS, ECS or serverless, and data stores sized for the workload you have rather than the one in the slide deck. I have worked in estates from a 15-account organization to hundreds of accounts serving more than 100 teams.
Why the bill moved, what each transaction or customer costs, and which five decisions shrink it. Commitments, rightsizing, allocation and the guardrails that stop surprises from coming back.
On-premises to AWS, version and engine upgrades, and moves to Graviton. Sequenced so that the risky step is small, reversible and rehearsed, with downtime measured in seconds where it matters.
Private connectivity between data centers and clouds. I have built it from inside the interconnection industry as well as from the customer side, so I know both ends of the cross-connect.
Kubernetes platforms, infrastructure as code and pipelines that developers actually like. Delivery measured with DORA metrics, not opinions.
Disaster recovery that has been tested, not assumed. Tiering, backup strategy, autoscaling, incident response and the post-mortems that make the next incident shorter.
Turning audit, pentest and assessor findings into a roadmap engineers can deliver. Evidence that is produced by the platform instead of by a scramble.
The infrastructure side of analytics: change data capture, streaming, lake and warehouse layers, orchestration, and the cost and access controls around them.
$ git log --oneline
Eighteen years across five kinds of environment. Each one taught something the others depend on.
Routing, firewalls, virtualization and server infrastructure, hands on, for manufacturing, services and international data center customers. It is where I learned that everything is eventually a networking problem, and that the person who understands the packets wins the argument.
Years inside the interconnection industry, working with customers alongside sales and project teams: designing and validating AWS architectures, connecting on-premises environments to every major cloud, and delivering migrations. It is where I learned to explain a design to the person paying for it.
AWS estates of hundreds of accounts: a group of production, development, staging and sandbox accounts for each of more than 100 teams, plus shared services and network accounts. Reusable Terraform modules, CI/CD with security scans in the pipeline, centralized monitoring, and coaching squads on DevOps practice. Scale and regulation leave no room for manual steps.
End-to-end ownership of AWS platforms in regulated US fintech: multi-account organizations covering production, non-production, security, audit and logging, workloads on EKS, infrastructure as code, FinOps for cloud and data, disaster recovery, and the platform's part in every audit. It is where architecture, the bill and the auditor all meet.
Leading cloud, DevOps and data engineering teams through team leads: hiring, performance reviews, roadmaps, budgets and reporting to senior leadership. I still read the Cost and Usage Report myself.
$ cat credentials.yaml
Amazon Web Services · valid through September 2027
The most demanding AWS architecture exam: complex multi-account design, migrations, cost control and continuous improvement.
Amazon Web Services · valid through August 2027
Previously held: AWS Certified DevOps Engineer – Professional.
$ echo "your problem" | mail fabio
A few lines are enough: what you run on AWS, what worries you, and what a good outcome looks like. I read everything, and I am always glad to compare notes.
Based in São Paulo, working remotely with teams in the Americas and Europe.
$ ls ~/projects