I build and operate cloud-native infrastructure that teams can trust — and lately, the platforms that put machine learning and LLMs into production. Thirteen years across big tech, investment banking, 5G telco and AI products: from moving a global telco’s core billing off the mainframe, to owning ML lifecycles end to end, to running agentic AIOps in production.
ROLE
DevOps / Platform / SRE — Senior
EXPERIENCE
13+ years
FOCUS
Platform · MLOps · AIOps · GenAI
CLOUDS
AWS · Google Cloud · Azure
CURRENTLY
System Engineer @ Apple
LOCATION
Brazil — remote worldwide
LANGUAGES
PT native · EN professional
STATUS
Open to opportunities
$ cat expertise.md
# What I do
Four areas where I go deep. The first two are where most DevOps engineers live; the last two are where I spend my time now.
Platform Engineering
Building the paved road so other teams ship without filing tickets.
Reproducible, auditable environments with Terraform, Ansible and cloud-init
GitOps delivery with Flux and Argo CD reconciling apps on Kubernetes
CI/CD across GitLab CI, GitHub Actions and Jenkins
Self-service tooling that removes humans from the critical path
Terraform
Kubernetes
Flux
Argo CD
GitLab CI
Helm
Reliability & Observability
Instrumenting systems before the incident, not during it.
Prometheus, Grafana, Alertmanager and OpenTelemetry distributed tracing
High availability and disaster recovery on life-critical systems
Incident response and postmortems that change the system, not the person
Cost optimization as a reliability concern, not an afterthought
Prometheus
Grafana
OpenTelemetry
AWS X-Ray
Sentry
ELK
MLOps
Owning the path from a model that works to a model that serves.
End-to-end ML lifecycle: data processing, training, tuning, deployment
AWS SageMaker pipelines and AWS IoT Greengrass for edge inference
Computer vision in production — PyTorch, YOLOv8, OpenCV
Treating models as deployable artifacts with the same rigor as services
SageMaker
PyTorch
YOLOv8
OpenCV
IoT Greengrass
AIOps & GenAI
Agentic workflows wired into CI/CD and monitoring — running in production today.
Agentic AIOps on the Anthropic SDK that investigates alerts automatically
Proactive issue detection ahead of the pager
Leading LLM adoption across the SDLC — skills, subagents and hooks
RAG-oriented design and LLMOps practices for embedding models in products
Anthropic Claude SDK
LLMOps
RAG
Python
Agentic CI/CD
$ grep -r "impact" ./experience
# Impact in production
Three projects where the number is the story. All of them ran in production, under real constraints.
Vodafone
The UK’s 999 emergency service, re-engineered for 5G
~1,000–2,000 ms
→
<100 ms
Emergency-call re-routing time
<10 s
AML (Advanced Mobile Location) response
A life-critical, heavily regulated system where a slow re-route is a person waiting for an ambulance. I designed the end-to-end Kubernetes architecture — non-production and production — and containerized the legacy stack, writing every manifest and a Helm chart per microservice.
Built the CI/CD pipelines and the whole delivery toolchain from scratch — Jenkins, Nexus, GitLab and Argo CD — plus observability and distributed tracing with Prometheus, Grafana, Alertmanager and OpenTelemetry.
Kubernetes
Helm
Argo CD
OpenTelemetry
5G
Regulated
Vodafone
Self-service VM provisioning for the whole engineering org
~2 weeks
→
instant
Time to provision, after approval
Provisioning a virtual machine meant a ticket and roughly two weeks of waiting. I designed and built the platform that removed the queue entirely.
A Node.js/Angular portal in front of Terraform, Jenkins, cloud-init and Python automation — approval stayed a human decision, everything after it stopped being one.
Terraform
Jenkins
cloud-init
Node.js
Self-service
IBM
A global telco’s billing core, off the mainframe
Mainframe / COBOL
→
Linux / Java
Core billing platform
600+
Red Hat Linux servers operated on vSphere
The largest billing application of a North American global telco, running on Mainframe and COBOL. I worked the migration to Linux and Java while operating the estate it ran on.
Production and development estates on vSphere alongside Mainframe MVS, with Jenkins, Docker, Terraform and Chef for delivery and configuration.
Linux
Java
COBOL
vSphere
Terraform
Chef
$ cat ~/.config/principles.toml
# How I work
automate_over_toil
=
true
If it hurt twice, it becomes a pipeline. Manual work is a bug report about the platform.
infrastructure_as_code
=
always
Nothing reaches production unless it is versioned, reviewable and reproducible.
observability_first
=
true
Instrument before the incident. Debugging blind at 3 a.m. is a choice made months earlier.
own_code_and_infra
=
true
Horizontal teams work. Whoever writes the service should be able to operate it.
blameless_postmortems
=
true
The failure is in the system, not the person. Otherwise people stop reporting the near-misses.
ai_as_leverage
=
true
LLMs take repetitive work off the path. They do not replace judgement, and I do not pretend otherwise.
$ git log --author="Felipe Miranda"
# Experience
System Engineer @Apple
Jun 2025 – Present·Cupertino, CA · Remote
Building and operating internal cloud-native platforms at the intersection of DevOps, Cloud and applied Generative AI, on a horizontal team where every engineer owns both code and infrastructure.
Build and operate the AWS infrastructure powering internal platforms and services; manage IaC with Terraform for reproducible, auditable environments.
Design and run CI/CD across GitLab CI and GitHub Actions with GitOps-based continuous delivery (Flux) reconciling applications on Kubernetes (EKS).
Implement monitoring and observability — AWS Managed Prometheus & Grafana, X-Ray, Sentry — to keep business-critical systems reliable and highly available.
Lead LLM / GenAI adoption across the SDLC, creating and sharing Claude skills, subagents and hooks; act as the team’s LLM/GenAI specialist for embedding LLMs into production.
Pioneer AIOps with agentic workflows on the Anthropic SDK integrated into CI/CD and monitoring, automating alert investigation and proactive issue detection.
Stack
AWS
Kubernetes (EKS)
Terraform
GitOps (Flux)
GitLab CI
GitHub Actions
Prometheus
Grafana
X-Ray
Sentry
Python
Django
Anthropic Claude SDK
DevOps / MLOps Engineer @CellarEye
Mar 2023 – Jun 2025·California, United States · Remote
Owned the path from model to production on a Computer Vision team building an AI-powered app that digitizes wine-cellar management (900+ bottles per collector).
Built the end-to-end ML lifecycle — data processing in PyTorch, object detection with YOLOv8, image processing with OpenCV — for real-time bottle recognition and cellar mapping.
Implemented MLOps on AWS SageMaker (training, tuning, deployment) and AWS IoT Greengrass for edge / on-device inference.
Designed CI/CD and managed Kubernetes on AWS for backend and iOS delivery; ran the infrastructure for scalability, security, high availability and disaster recovery.
Stack
Python
PyTorch
YOLOv8
OpenCV
AWS SageMaker
IoT Greengrass
Kubernetes
CI/CD
MLOps
Computer Vision
Edge
iOS
DevOps Specialist @Vodafone
Apr 2021 – Mar 2023·Portugal
Migrated the UK’s 999 emergency-call service to 5G — re-engineering and containerizing a life-critical, highly regulated system on Kubernetes.
Designed the end-to-end Kubernetes architecture (non-prod + production) for high availability and fault tolerance; containerized legacy apps with all manifests and a Helm chart per microservice.
Built the CI/CD pipelines and delivery toolchain from scratch (Jenkins, Nexus, GitLab, Argo CD / GitOps); implemented observability and distributed tracing (Prometheus, Grafana, Alertmanager, OpenTelemetry).
Designed and built a self-service VM provisioning platform — Node.js/Angular portal on top of Terraform, Jenkins, cloud-init and Python.
Impact Cut emergency-call re-routing from ~1,000–2,000 ms to under 100 ms, brought AML response under 10 seconds, and reduced VM provisioning from ~2 weeks to instant after approval.
Stack
Kubernetes
Helm
Terraform
Jenkins
GitLab
Argo CD
cloud-init
Python
Prometheus
Grafana
OpenTelemetry
Elasticsearch
Redis
vSphere
DevOps Engineer @TempoTem
Jun 2020 – Apr 2021·Barueri, São Paulo, Brazil
Built and operated high-availability, 100% AWS cloud-native e-commerce environments.
Ran the full AWS estate — EC2, S3, RDS, DynamoDB, Route 53, Lambda, CloudFront, CloudFormation and EKS — with Jenkins, Bitbucket, Prometheus, Grafana and Magento.
Stack
AWS
EKS
CloudFormation
Jenkins
Bitbucket
Prometheus
Grafana
Magento
DevOps Engineer @Veloe
Feb 2019 – Jun 2020·Barueri, São Paulo, Brazil
Operated a 100% AWS production environment for Java, Node.js, React and Python applications.
Delivery and monitoring with Azure DevOps, EKS, Docker, CloudFormation, Dynatrace and CloudFront.
Stack
AWS
EKS
Docker
Azure DevOps
CloudFormation
Dynatrace
CloudFront
DevOps Engineer @BTG Pactual
May 2018 – Feb 2019·São Paulo, Brazil
Introduced DevOps culture and built architectures for high-availability internal systems at one of Latin America’s largest investment banks; 100% AWS cloud-native.
AWS (EC2, S3, RDS, DynamoDB, Route 53, Lambda, CloudFront, CloudFormation, EKS) with Jenkins, GitHub Enterprise, Prometheus and Grafana.
Stack
AWS
EKS
CloudFormation
Jenkins
GitHub Enterprise
Prometheus
Grafana
DevOps Engineer @IBM
Aug 2013 – Jul 2017·Hortolândia, São Paulo, Brazil
Operated 600+ Red Hat Linux servers on vSphere (production and development) plus Mainframe MVS for a large global telco client.
Key project: migrated the largest billing application of a North American global telco from Mainframe/COBOL to Linux/Java.
Delivery and configuration with Jenkins, Docker, Terraform and Chef; Java and Micro Focus COBOL.
Stack
Red Hat Linux
vSphere
Mainframe MVS
Jenkins
Docker
Terraform
Chef
Java
COBOL
System Administrator @Earlier experience
2008 – 2013·Brazil / Portugal
Manzano Pneumáticos, Concrete Solutions, Novo Motor and Credimais — Linux server administration, networking and firewall (Squid proxy, VPN, Active Directory), Apache/Magento performance tuning, and web-portal deployments on AWS EC2.
Stack
Linux
Apache
Squid
VPN
Active Directory
AWS EC2
$ ls -1 ./skills
# Skills
Cloud
AWS
EKS
EC2
RDS
MSK
OpenSearch
ElastiCache
SageMaker
IoT Greengrass
CloudWatch
X-Ray
Google Cloud
Azure
Containers & Orchestration
Kubernetes
Helm
Docker
OpenShift
Infrastructure as Code
Terraform
Ansible
CloudFormation
cloud-init
CI/CD & GitOps
GitLab CI
GitHub Actions
Jenkins
Argo CD
Flux
Observability
Prometheus
Grafana
Alertmanager
OpenTelemetry
AWS X-Ray
Sentry
ELK / Elasticsearch
MLOps / AIOps / GenAI
AWS SageMaker
PyTorch
Model deployment
Agentic workflows (Anthropic Claude SDK)
LLMOps
RAG-oriented design
Programming
Python
Bash
JavaScript / Node.js
Databases
PostgreSQL / RDS
DynamoDB
Redis
Elasticsearch
MongoDB
Practices
SRE
High availability
Disaster recovery
Incident response
Cost optimization
$ ls ./certifications
# Certifications
AI Fluency for Builders — Anthropic
AI Fluency for Students — Anthropic
Teaching the AI Fluency Framework
AI Capabilities and Limitations
Introduction to Claude Cowork
Big Data — Fundamentals
15 certifications in total — full list available on LinkedIn.
$ cat education.md
# Education
Computer Science
IBTA — Instituto Brasileiro de Tecnologia Avançada