Site reliability engineer · Lisbon
So engineers can build, products can scale, and on-call doesn't have to be chaos.
11+ years building and operating production systems at scale — from Dailymotion's 300M+ monthly users and 17+ Kubernetes clusters to OLX.
I turn infrastructure complexity into reliable platforms, repeatable automation, and observability that drives decisions.
Now I'm bringing AI agents into operations — reducing toil, accelerating incident response, and keeping humans in control.
Clusters, pipelines, dashboards and pages at 3am. Good infrastructure is the part of a product you never notice.
Eight roles since 2012. Four systems of my own.
2012 → now
Open any tile for details
Record
SRE on a platform team. Owned the infrastructure-as-code stack (Terraform, Terragrunt, Atlantis) and built a tool that explains every Terraform plan in plain English, so developers catch mistakes before they apply; it cut plan review time by about two-thirds. Built an event-processing platform for internal tooling that ingests more than 50k messages a day, and benchmarked and introduced AI tooling into the SRE workflow.
Blockchain RPC infrastructure. Cost and performance work on the node fleet, real-time monitoring and alerting, and automated provisioning with health-check-driven failover so incidents resolved themselves, cutting time to mitigate by about 23%.
Kubernetes at production scale across regions and cloud providers. CI/CD migration to JenkinsX, Terraform and Atlantis standardised across engineering, and a centralised observability rollout on Prometheus, Grafana and Loki.
A multi-region AWS footprint of 400+ virtual machines across the US, Europe and APAC. Automation with Terraform and Ansible, capacity planning for high-traffic video.
Early DevOps hire on a cloud telephony platform: infrastructure redesign with autoscaling, containerised development workflows and a first monitoring stack.
A legacy monolith with its own servers moved to microservices on AWS, including the SQL database move to RDS.
Built the agency's infrastructure from scratch: hosting, mail and client-facing services.
Before infrastructure: partner programmes, physical and online, built and measured with the project director.
Work of my own
Active
Run AI agents like infrastructure. Not like experiments.
A Git-native control plane for managing distributed, self-hosted AI agents across machines.
One Go binary. One declarative YAML manifest. Git as the single source of truth.
Kuby turns a scattered collection of agents into an observable fleet, with centralized inventory, execution history, metrics, and operational visibility, without a configuration database or another dashboard full of settings.
Active
When production breaks, clarity is everything.
An AI-powered mission control for SREs, built to turn operational chaos into actionable intelligence.
From incident triage and AWS dependency mapping to multi-agent code reviews, Warroom brings investigation into a single terminal-first workflow.
Specialized agents connect the signals, trace the impact, and surface evidence-backed next steps.
Agents investigate. Engineers decide.
Active
From legal documents to actionable intelligence. At scale.
An asynchronous, auditable AI processing platform built to turn complex document analysis into a reliable product experience.
Built to orchestrate everything from document ingestion and asynchronous AI processing to notifications and payment conversion.
Under the hood: queue-driven workloads, comprehensive audit trails, and a flexible architecture designed to make complex document processing reliable, traceable, and scalable.
Active
Production traffic. Under control. Even when AI takes the wheel.
A lightweight, agentless control plane for managing Traefik routing across remote environments — all from a single terminal.
Built in Go, it brings infrastructure-grade safety to traffic management: change planning, drift detection, automatic backups, and rollback. No agents to deploy. No additional infrastructure to maintain.
AI agents operate through tightly scoped MCP tools, with dry runs and explicit policies controlling which routes and domains they can modify.
Every change is intentional. Every action has boundaries. Every mistake has a way back.
Scale
Dailymotion, 2017 to 2023. *Self-reported, platform-wide for the period; the incident figure is attributed to the centralised observability rollout.
Contact