Senior Site Reliability Engineer in Bremen. I build and run Kubernetes platforms on AWS that stay up.
From Nairobi to Bremen, from writing the software to keeping it running. One chapter per screen.
$ whoami dennis $ cat degree.txt Computer Technology in Computer Science The Technical University of Kenya Honours: First Class Sep '15 — Sep '19
Computer Technology at The Technical University of Kenya, graduating with First-Class Honours.
$ php artisan migrate Migrating: create_members_table Migrated $ npm run build ✓ vue app built $ git push origin master
Backend APIs and web apps in PHP, Laravel and Vue.js for a community platform and a company intranet, built with pair and mob programming.
$ flask db upgrade ✓ schema up to date $ aws lambda update-function-code \ --function-name api --zip-file fileb://api.zip "LastUpdateStatus": "Successful" $ curl -s localhost:5000/health {"status": "ok"}
Ran the AWS architecture behind a diversity-hiring job board and led its backend API. Before that, a React Native app for small-scale farmers.
$ terraform apply aws_ecs_service.api: Modifying... aws_ecs_service.api: Modifications complete Apply complete! Resources: 0 added, 3 changed $ docker push $ECR/api:latest latest: digest: sha256:9f2c… size: 2417
Led the team and the AWS platform for a training-course marketplace for career changers: Terraform, Docker, ECS and CI/CD.
$ terraform workspace select prod $ terraform apply -target=module.ecs Apply complete! $ aws ecs list-services --cluster prod web-de web-en api $ echo "team: 5 engineers"
Led five engineers building a citizen-participation platform for energy projects in three languages, used by the City of Bremerhaven and Landkreis Vechta.
$ argocd app sync prod-api Sync Status: Synced Health: Healthy $ kubectl -n prod get pods NAME READY STATUS api-7d9f6c-x2 1/1 Running api-7d9f6c-q8 1/1 Running $ heroku ps:scale web=0 -a legacy
Moved the whole platform from Heroku to Amazon EKS with near-zero downtime. GitOps with Argo CD, Terraform and Helm, at a much lower bill.
$ kubectl get nodepools NAME NODECLASS NODES READY default default 12 True $ holmes ask "why did checkout latency spike?" → correlating Mimir · Loki · Tempo → root cause: db pool exhausted at 14:02
Own the company-wide Grafana LGTM observability platform, Karpenter autoscaling, an AI troubleshooting assistant on HolmesGPT, and incident command.
Production EKS clusters from bootstrap to day two, with node capacity that follows the workload instead of a guess.
Accounts, networks and compute laid down as code, reviewed like code, and right-sized so the bill stays sane.
Pipelines that turn a merge into a safe rollout, with Git as the source of truth for what runs in dev, staging and prod.
Metrics, logs, traces and profiles from every cluster and account, correlated in one place and wired to SLOs.
Incident commander on call: mitigate first, keep stakeholders informed, then a postmortem that changes something.
The glue: tooling, operators and scripts that remove a manual step for good.
A Kubernetes platform on AWS your teams can ship to on their own: clusters, GitOps pipelines, secrets and guardrails.
EKS · Terraform · Helm · Argo CDOne place for metrics, logs, traces and profiles across every cluster and account, with alerts that page only when users notice.
Grafana LGTM · OpenTelemetry · SLOsMove off a PaaS or hand-built servers onto containers, database and all, without a big-bang cutover.
Heroku → EKS · near-zero downtime