← Services

Services

Cloud architecture and MLOps on AWS, GCP and Azure

I design the infrastructure where data and models run: Kubernetes, infrastructure as code, CI/CD and environments that behave the same in development and production. I started in cloud infrastructure in 2013, so I also know how to read the bill.

Problems I solve

  • What works in the notebook or locally breaks in production.
  • The cloud bill keeps growing and nobody quite knows why.
  • Models are deployed by hand and cannot be reproduced.
  • You need to choose between infrastructure options and compare their real cost.

What I do

  • Kubernetes (EKS, GKE, AKS) with autoscaling and costs under control.
  • Infrastructure as code with Terraform, Pulumi and Helm.
  • MLOps: orchestration with Dagster or Airflow, model registry with MLflow.
  • Observability and total cost of ownership (TCO) analysis.

Related projects

How I work

  1. 01 Diagnosis. I understand the problem, the data and the constraints, and we check whether it makes technical and economic sense.
  2. 02 Prototype. Something working with real data in a few weeks, so decisions rest on evidence.
  3. 03 Implementation. We take it to production with your team, with tests, monitoring and clear costs.
  4. 04 Handover. Documentation and training so the team can keep going without depending on me.

I work per project, as a monthly advisor or as an interim (fractional) data and AI lead. On-site in Santiago and Antofagasta, and remote for the rest of Chile and abroad.

FAQ

Which clouds do you work with?
AWS, GCP and Azure. The choice depends on what you already have, your team and cost, not on my preference.
Do I need Kubernetes?
Not always. Sometimes a managed service is simpler and cheaper. We evaluate it with numbers before deciding.