Work
Problems, constraints, and what actually changed
Most of this work sits under NDA, so these describe the sector and the scale rather than the client. Each one leads with the constraint — that's the part that made it hard, and the part worth reading.
Infrastructure02
SaaS platform, EU · 2025
Replacing the network under a running production fleet
A staged migration to Cilium and eBPF across a live cluster fleet — kernel-level network policy, no customer-visible downtime, and traffic visibility the team had never had.
- Cilium
- eBPF
- Hubble
- Kubernetes
- Terraform
- Gateway API
- Downtime
- None
- Node pools migrated
- 9
- Unknown dependencies found
- 23
B2B SaaS platform · 2024–2025
From one cluster per customer to one platform
A SaaS product was scaling infrastructure cost linearly with customers. I designed the tenancy model that broke that curve — real isolation, per-tenant cost visibility, and an escape hatch for the customers who genuinely needed one.
- Kubernetes
- Kyverno
- Cilium
- Argo CD
- Terraform
- OpenCost
- Clusters
- 40 → 6
- Onboarding time
- 3 days → 12 min
- Infra cost per customer
- −64%
These are written up at length because the reasoning is the point. If you'd rather see the whole span of the work — every project since 2014, in a fixed format, filterable by discipline — the portfolio archive is the page for that.
08Next step
Recognise your own problem in one of these?
Most engagements start with a short call and a written plan — what I'll do, what it costs, and what done looks like.
Stockholm, Sweden (CET) · Comfortable overlap with European hours and US East Coast mornings
- 01
You write
A paragraph is enough. What's breaking, or what you're trying to build.
- 02
We talk
Thirty minutes, no charge. I'll tell you what I'd look at first.
- 03
You get a scope
Fixed price, fixed deliverable, written down before anyone commits.