now
Leading cloud infrastructure at Protex AI, a SaaS scale-up
Setting technical direction, reviewing the important pull requests, and mentoring engineers. Incident response when production burns. Still writes infrastructure code every week.
$ whoami
Gergely Peidl
Lead DevOps & Cloud Infrastructure Engineer. Thirteen-plus years of making servers do useful things. I build cloud platforms that stay boring, and I automate everything that doesn't.
~/about
I turn runbooks into pipelines
The short version of thirteen years.
I'm a lead DevOps and cloud infrastructure engineer based in Budapest. I started out as a full stack software engineer, mostly on the frontend side, and slowly moved toward infrastructure. Somewhere along the way I found out that I enjoy making servers boring.
These days I lead a cloud infrastructure team: setting technical direction, reviewing the important changes, and still writing infrastructure code every week. I care about infrastructure as code done properly, monitoring people actually look at, and automation as a default habit instead of an afterthought. My favourite tool of recent years is Terragrunt. It keeps large Terraform setups organized, and I will happily talk about it longer than you planned.
There is one rule behind all of it: if something breaks once, I make sure it never breaks that way again.
“Talk is cheap. Show me the code.”
- role
- Lead DevOps & Cloud Infrastructure Engineer
- base
- Budapest, Hungary · CET/CEST
- focus
- AWS · Kubernetes · IaC · CI/CD · Automation
- superpowers
- Transparent communication with all stakeholders
- Turning repeated work into pipelines
- certificates
- AWS Certified Cloud Practitioner · IELTS C1
- interviewing
- 70+ technical and hiring manager interviews conducted
- off the clock
- Bouldering, DIY electronics, mentoring & guest lectures
~/experience
Thirteen years, two worlds
A project-based consulting company and a SaaS scale-up. They teach very different lessons, and both show up in how I build.
along the way
DevOps tech lead at Sonrisa, a project-based consulting company
Twelve years of client work, from on-prem data centers to a PCI DSS compliant AWS identity platform serving an organization of 300+ engineers. Built the CI/CD practice from Jenkins to cloud-native, moved estates to AWS, and shipped 500+ production deployments.
in the beginning
Full stack software engineering
Mostly frontend, back when browser compatibility was a real sport. Side quest that never ended: Linux. I use Arch, btw.
// comfort zones
AWS OpenTofu / Terragrunt Ansible ArgoCD Datadog / Snyk / Aikido Cilium Kyverno Trivy Velero Authentik / SOPS Home Assistant / ESPHome Python / Bash// what former colleagues say
Professionals like him are rare to find. He takes ownership of initiatives from start to finish, ensures they are executed flawlessly, and always speaks up with insightful opinions.
Every piece of advice he gave turned out to be spot on. Thanks to his inspiration, I even started building my own homelab.
~/stack
Tools I reach for
No percentage bars. Either I can debug it at 3 AM or I can't.
Cloud & Platform
- AWS: ECS, VPC, IAM, RDS & friends
- Cloudflare for DNS, CDN and edge stuff
- Proxmox VE for bare-metal virtualization
- Linux: Arch as my daily driver
IaC & GitOps
- OpenTofu / Terraform
- Terragrunt for sane multi-stack layouts
- Ansible for everything with an SSH port
- ArgoCD + Renovate for pull-based ops
Containers & Orchestration
- Kubernetes: my active deep-dive, Talos-flavored
- Docker / Podman
- Helm / Kustomize
- Gateway API over vintage Ingress
Observability
- Datadog: production telemetry at the day job
- kube-prometheus-stack, Grafana, Loki, Tempo in the lab
- OpenTelemetry instrumentation
- Alertmanager → Signal, because SMS is not a strategy
Security & Policy
- Snyk + Aikido: dependency scanning and AppSec programs
- SOPS + age for secrets in git
- Kyverno policies, enforced not decorative
- Trivy for CVE scanning
- Cilium NetworkPolicy defaults-deny habits
Automation & IoT
- Python & Bash glue, industrial grade
- Home Assistant as household kernel
- ESPHome firmware on self-built devices
- n8n workflows, 3D-printed enclosures
~/homelab
planet-express
“Good news, everyone!” A production-grade Kubernetes homelab on a single bare-metal host, run entirely as code, and where I'm learning Kubernetes properly, bottom-up. Every non-obvious choice has an Architecture Decision Record that explains what was evaluated, what lost, and why.
22architecture decisions on record
4-nodeTalos Kubernetes cluster
100%IaC
€0monthly cloud bill
// case studies
~/automation
Home Automation & IoT
Where the homelab philosophy gets a roof, walls and Wi-Fi.
Home Assistant has run our home since 2017. The fleet passed fifty devices a while ago: self-built ESPHome boards soldered, flashed and boxed in 3D-printed cases, plus off-the-shelf gadgets that behave (local-first, MQTT or Zigbee, no cloud account).
Jenkins, retired
My home jobs used to run on a Jenkins VM. Today the same things run as Home Assistant automations and n8n workflows: lighter, simpler to maintain, no plugin updates on a Sunday.
Upstream contributions
Documentation fixes and Hungarian translations for Home Assistant, listed on the official credits page. Small diffs, real project.
Environmental fleet
Temperature, humidity, pressure and CO₂ sensors in every room. Comfort is much easier to tune when you can see the numbers.
Energy everywhere
Plug-level power monitoring on everything that draws current. The dishwasher claims eco mode. The meter disagrees.
Presence-aware lighting
Lights that adjust themselves before anyone looks for a switch, driven by real presence instead of motion timers.
ESPHome everywhere
Every DIY sensor runs ESPHome firmware. YAML in, measurements out, updates over Wi-Fi.
DIY hardware
ESP32 and ESP8266 boards, a soldering iron and a 3D printer. If it exists, it can probably carry a sensor.
Fail-safes first
Every automation has a manual override and sane fallbacks. The house serves its humans, not the other way round.
Empty house protocol
When all our loved ones, including the dog, are out, presence detection starts the robot vacuum. Loud cleaning belongs to an empty flat.
Automate the friction away, but always keep a manual override. Every guest should still find a working light switch.
~/contact
Say hello
Open to interesting problems, mentoring conversations, or a good debate about whether YAML counts as programming.