Available for new engagements
I build the infrastructure layer — and I get called when it has to hold up against something hard.
Twelve years in cloud infrastructure. Production Kubernetes platforms, multi-cloud architecture, migrations that stay boring, and data infrastructure run with the same discipline as everything else. I design it, build it, and leave your team able to run it without me.
Anshuman Abhishek
Platform Engineering Architect
Stockholm, Sweden · CET · 12 years
Working across
- Experience
- 12 yrs
- Clouds
- 3
- Certifications
- 5
- Typical utilisation found
- 35–50%
In production infrastructure since 2014, from Linux estates to multi-cloud platforms.
AWS, Azure and Google Cloud — plus hybrid and bare metal when that's the right answer.
CKAD, GCP Professional Cloud Architect, AWS SA, Azure AZ-104, Terraform Associate.
What most Kubernetes clusters are actually running at when I first open them.
Verified credentials
Certified across every major cloud
Certs are a floor, not the pitch — twelve years in production is the real argument. But procurement checks, so here they are: five badges, every link verifiable.
What I do
Six things I'm hired for
They're the same job seen from different angles. Every one of them is infrastructure that has to survive a real constraint — scale, regulation, cost, or a deadline.
Clusters your team can actually run
Platform & Kubernetes
Production and multi-tenant Kubernetes, internal developer platforms with self-service golden paths, and the cloud-native networking underneath. The part most teams underestimate is the network and the tenancy model — that's where I'm most useful.
- Kubernetes
- Cilium
- eBPF
- Istio
- Gateway API
Migration day should be boring
Cloud Architecture & Migration
Landing zone design and workload migration across AWS, Azure and GCP — and, increasingly, the return trip. Most migrations fail in the planning, not the execution, so that's where I spend the time.
- AWS
- Azure
- GCP
- Terraform
- OpenTofu
Data infrastructure, run like a platform
Data Platform Engineering
I don't build your dbt models — your analytics engineers are better at that. I make sure the platform underneath them is reliable, versioned, observable and not quietly costing a fortune.
- Airflow
- Dagster
- Kafka
- Strimzi
- Spark
Turn tribal knowledge into reviewed code
Automation & Infrastructure as Code
The pattern is always the same: one person knows how to do the thing, it takes three hours, it happens twice a week, and it breaks differently every time. That's a design problem, and the fix is usually smaller than people expect.
- Terraform
- OpenTofu
- Ansible
- Python
- Go
Everyone has a DR plan. Almost nobody has a tested one.
Reliability & Resilience
The boring half of infrastructure — the half that only matters on the worst day of the year, which is exactly why it never gets prioritised until after that day has happened.
- Prometheus
- Grafana
- Loki
- OpenTelemetry
- LitmusChaos
Between the people who can't implement and the people who can't read a regulation
Security & EU Compliance
Compliance consultants produce documents. Infrastructure engineers produce systems. What's usually missing is the person who can take an obligation and turn it into Terraform, an admission policy and an audit trail that survives an assessor.
- Kyverno
- OPA
- Vault
- Sigstore
- Trivy
Selected work
What the work actually looks like
Client names are withheld under NDA, so these describe the sector and the scale instead. Each one leads with the constraint, because that's the part that made it hard.
How I work
The method, not the mystique
Most architecture arguments are really disagreements about unstated requirements. Most migrations fail in the planning. This is how I avoid both.
01
Discovery before design
I start with what actually exists rather than what the documentation claims. Real dependencies, real traffic, real costs — and a written list of the questions nobody could answer. That list is usually the most valuable page.
02
Options, then a recommendation
Every significant decision gets two or three candidates with honest trade-offs and a recommendation with reasons, recorded as an architecture decision record. One option is a preference. Three with a recommendation is judgement.
03
Build with the failure condition defined
Rollback criteria on migrations, success criteria on proofs of concept, abort conditions on cutovers — written down before work starts, not improvised on the day.
04
Hand over properly
Reference implementations, runbooks, recorded walkthroughs. The measure of the work is whether it still runs well six months after I leave.
Stack
What I work with
Breadth is the point. You cannot weigh three options honestly, or reason about where the second bottleneck is, unless you've seen a lot of systems.
- Cloud
- AWS · Azure · Google Cloud · IBM Cloud · Hybrid & on-premise
- Kubernetes
- EKS / AKS / GKE · OpenShift · Self-managed · Operators & CRDs · Helm · Kustomize · Karpenter
- Networking
- Cilium / eBPF · Calico · Istio · Linkerd · Gateway API · Envoy · NGINX
- Infrastructure as Code
- Terraform · OpenTofu · Crossplane · CloudFormation · Ansible · Packer
- Delivery
- Argo CD · Flux · GitHub Actions · GitLab CI · Jenkins · Argo Rollouts · Tekton
- Data
- Airflow · Dagster · Kafka / Strimzi · Spark · Snowflake · Databricks · BigQuery
- Observability
- Prometheus · Grafana · Loki · Tempo · Thanos · OpenTelemetry · ELK
- Security
- Kyverno · OPA · Vault · Sigstore · Trivy · Falco · NIS2 · DORA
- AI Infrastructure
- Amazon Bedrock · Kueue · KServe · vLLM · Ray · GPU Operator
- Languages
- Python · Bash · Go · TypeScript · Linux
Writing
Notes from the work
Mostly the things I had to figure out the hard way.
- Mar 2024· 2 minImplementing Security on Azure DevOps Pipelines
- Aug 2022· 4 minWhy Service Mesh Is Necessary Today
- Jun 2022· 16 minTerraform: Production-Grade IaC Coding
Teaching
I also teach this. Badly-taught DevOps is why so much of it is done badly.
Most courses teach you to run commands. Almost none teach you to make the decision — why this CNI, why this tenancy model, what you give up when you pick the easy option. That's the part that takes a decade to learn and about a day to explain properly.
Training, mentoring and workshops1:1 mentoring
For engineers moving from DevOps into platform or architecture roles. Real systems, real reviews, not exercises.
Team workshops
Kubernetes, GitOps, Terraform, cloud-native networking — taught against your actual stack rather than a lab.
Architecture coaching
For teams who have the skills but keep making decisions they regret six months later.
Tell me what's broken.
Or what you're trying to build. I'll give you an honest read on it before either of us commits to anything — and if I don't think I'm the right person for the job, I'll say so on that call.
Stockholm, Sweden (CET) · Comfortable overlap with European hours and US East Coast mornings