01
Cloud & systems architecture
Right-sized environments: private networks, load balancers, autoscaling, and Linux hosts tuned to the workload instead of the tutorial.
Available for new projects
Infrastructure that's boring by design.
I'm Otto — a systems and cloud infrastructure engineer. I design, automate, and operate Linux environments that stay quiet, even when traffic doesn't.
Linux
Nginx
Docker
Kubernetes
Terraform
Python
Go
PostgreSQL
Redis
Prometheus
Grafana
GitHub Actions
Cloudflare
Ansible
systemd
WireGuard
01 Services
What I do, and when it helps.
Four disciplines, usually in combination. No dashboards nobody reads, and no architecture diagrams that never ship.
02
Automation & delivery
CI/CD pipelines, declarative configuration, and small Go or Python tools, so a deploy is one command and a rollback is one flag.
03
Observability & reliability
Metrics, logs, traces, and alerting that pages only for real problems. SLOs agreed before the incident, not reconstructed after it.
04
Hardening & access
Least-privilege access, hardened hosts, TLS everywhere, secrets out of the repository, and patching that actually happens.
Reference shape
small surface area · boring failure modes
01
Edge
Cloudflare · TLS · WAF
02
Proxy
Nginx · routing · cache
03
Runtime
Go · Python services
04
Data
PostgreSQL · Redis
02 Selected work
Measured by what changed.
A few engagements, described by outcomes rather than tools installed. Client names withheld under NDA — details available on request.
01 A monolith, containerised with zero cutover downtime 14 services moved off hand-managed VMs onto Docker and CI, with a staged blue/green cutover and written runbooks. 0 mindowntime −41%infra spend 3×deploys
Scope
A 14-service monolith spread across six hand-tended VMs, deployed by SSH and hope, with no staging environment.
Approach
Containerised one service at a time behind a reverse proxy, then moved traffic in stages. Each stage had a tested reverse in under two minutes.
Outcome
Cutover with no downtime, 41% lower hosting spend, and deploys that moved from monthly to several times a week.
02 Multi-region failover for an API that couldn't go down Two regions, replicated data, health-checked routing — recovery measured in minutes instead of hours. 4 minRTO 99.98%uptime 48 msp95
Scope
A single-region Python API serving EU customers, backed by a four-hour manual recovery procedure that nobody had tested.
Approach
Promotable read replicas, DNS failover wired to health checks, and a scheduled restore test that ran whether or not anyone was watching.
Outcome
A documented four-minute recovery, p95 down from 210 ms to 48 ms after query and cache work, and drills that stopped being scary.
03 An Nginx edge layer that shrank the blast radius Caching, rate limiting, and a tuned TLS path for 40M monthly requests — origin load down by nearly two thirds. −62%origin load A+TLS grade 40Mreq / month
Scope
Origin servers saturating during traffic spikes, with no cache layer and a default Nginx configuration carrying production traffic.
Approach
Edge caching with surgical cache keys, per-route rate limits, HTTP/3, OCSP stapling, and a config split into auditable includes.
Outcome
62% less origin traffic, no 5xx through peak events, and TLS graded A+ across every endpoint.
04 CI/CD rebuilt around audited, reversible deploys Nine engineers, one pipeline, every release attributable and one flag away from a rollback. 34→4min deploy 100%traceable 0secrets in git
Scope
Three competing deploy scripts, credentials committed to a repository, and no consistent trail of who shipped what.
Approach
A single pipeline with pinned images, signed artefacts, vaulted secrets, and a one-command rollback rehearsed by the team, not by me.
Outcome
Deploy time from 34 minutes to 4, every release attributable to a commit, and secrets permanently out of version control.
03 Process
Four steps, no surprises.
Short engagements, small surface area, no surprise invoices. Most work runs four to six weeks, start to handover.
01 Audit
Find out what's really running
Hosts, traffic paths, failure modes, and spend — mapped in two weeks. You keep a written findings document either way.
02 Design
Two options, priced and argued
A target architecture with the trade-offs written down, so the decision is made on facts rather than preference.
03 Build
Small, reversible steps
Changes land behind flags in reviewable increments. You see the diff, the plan, and the rollback before production does.
04 Operate
Handover that sticks
Runbooks, dashboards, alert routes, and a recorded session for whoever holds the pager next.
04 De-risking
You keep the keys, the docs, and the exit.
Infrastructure work goes wrong when it's opaque. So the scope gets written down, the changes stay reversible, and nothing ships without a way back.
Talk through your setup
Written scope before work
Deliverables, acceptance criteria, and a change process agreed in writing before anything starts.
Reversible by default
Every change ships with a documented rollback. No irreversible first steps at two in the morning.
Documentation is a deliverable
Runbooks, diagrams, and an access map live in your repository — not in my notes.
Handover is the finish line
Your team can operate it without me. That is the definition of done, not an upsell.
99.99%
Availability target
≤1day
Typical response time
100%
Documented handovers
0
Proprietary lock-in
Operating principle
If it isn't monitored,
it isn't finished.
If it can't be rolled back,
it isn't ready.
Otto — infrastructure engineering
05 Questions
Worth asking before you hire anyone.
Straight answers to the things that usually get skipped in a first call.
01What does a typical engagement cost?
Most projects land between three and six weeks of focused work, quoted as a fixed price after the audit — so you're not watching a meter run. Small, well-defined fixes are fine too.
02Do you work alongside an existing team?
Usually. I'd rather improve your pipeline than replace it. I pair with your engineers, write in the open, and leave the architectural decisions with you — I'll argue my case, then respect the call.
03How do you handle access and secrets?
Scoped, time-limited credentials through your own vault or password manager. No production secrets in shell history, tickets, or chat. Access is inventoried at the start and revoked, in writing, at handover.
04Do you offer ongoing support?
Yes — a light monthly retainer for patching, alert triage, and small changes, plus a quarterly review of what's drifting. Month to month, cancellable with 30 days' notice.
05What if we're not sure what we need?
Start with the audit. It's fixed-price, it produces a document you own, and it regularly ends with "you don't need me for this" — which is a fine outcome for both of us.
06Are you available for emergencies?
Sometimes. Existing clients get priority; new emergency work depends on the week. Ask anyway — worst case I'll point you at someone who can help faster.
06 — Contact
Let's make your infrastructure unremarkable.
Tell me what's breaking, what's slow, or what's about to. You'll get a short, technical reply within one business day — no pitch deck, no calendar maze.
- Telegram
- @OttO4G
- [email protected]
- Response
- Within 1 business day
- Hours
- 09:00–18:00 UTC
- Working
- Remote · EU & MENA