Senior Platform & Systems Engineer, Engineering Leader
Systems architect and hands-on engineering leader for AI and platform work. I design how infrastructure, platform and product fit together, make that design run, and lead the team that keeps it running.
I build systems a team can keep running after I leave, for companies the same way I first built them for myself: Kubernetes on its own hardware, GitOps and CI/CD that ships daily, Go services on durable workflows, private AI for documents and bookkeeping, and AI agents that work under rules, with typed tool policies, human approval before any change, single-owner lanes and evals against independent ground truth.
Engineering since 2012, including Engineering Manager at Homeday and CTO of an EXIST-funded startup: teams built, a QA department designed from zero, review and release practices that hold under growth. Since 2022 I run all of it as one self-hosted platform at Glowing Pixels.
I work two ways: as fractional CTO or engineering manager for a team that needs technical direction, process and hiring, and hands on when a platform has to be built, migrated or stabilised. The work fits when product, security and engineering see the same initiative differently, or when a pilot needs to become an operable system.
Experience
Venture exploration 2022 - 2024 (Antler); platform research and build 2025; adoption and offers 2026.
- Platform. Self-hosted and run as one system: forge, GitOps-managed Kubernetes clusters (Argo CD), secrets in Vault, PostgreSQL on CloudNativePG with point-in-time recovery, CI runners that KEDA scales on load; one merge policy for every repository, rollback in a single step; duplicated Go code extracted into shared kits the internal services build on.
- Documents. Automation for clients: classification, extraction and routing on private inference, so no document, prompt or model output leaves the client's network.
- Finance. Bookkeeping as software: bank, Stripe and PayPal transactions pulled through APIs, posted automatically, exported as DATEV batches, with tests pinned to the tax assessments. E-invoicing (XRechnung, ZUGFeRD); incoming invoices read by OCR and local models, every field with a confidence score and owner approval before posting.
- Workflows. Durable processes on Temporal in Go: mailbox sync, an ERP's dunning process, and the control plane of a coding-agent service on Kubernetes; idempotent steps, versioned signals, replay of recorded histories as a test.
- AI agents. Several coding agents ship into the same repositories under single-owner lanes, tiered escalation and verification rules written from real failures.
- Mobile and desktop. A Flutter app for Android and iOS with barcode scanning and its own server API, signed releases from an own pipeline; a native Android app in Kotlin (Jetpack Compose) with on-device models, speech recognition and OCR; a macOS menu-bar app in SwiftUI with Keychain and Touch ID; maintainer of the leading fork of the Dendrite Matrix homeserver.
- Teams. Grew one cross-functional team into two (five to eight engineers each); raised CI/CD, observability, test strategy and on-call until releases stopped being events.
- QA department. Chosen to build a company-wide QA department from scratch: negotiated what it owns and where it meets every other department, designed the quality workflow from ticket to release, ran the hiring loop, hired its first senior engineers.
- Outcome. Several teams adopted the practices before the property-market downturn halted the rollout.
- Led the company-wide documentation platform and docs infrastructure.
- Automated checks, builds and publishing for more than 300 projects across all products.
- Modernised service boundaries and deployment of an insurance platform; automated internal operations.
- Founded and built the engineering team; owned the architecture of a deep-learning product.
- Split the monolith into services, designed the core API, localised the platform into more than ten languages.
- Built the Berlin team from scratch across three locations as engineering grew from about 3 to 20 to 30 people.
- Rails back end, AngularJS front end, Ansible-based deployment.
- Led a multi-year, externally funded R&D programme in MEMS sensing and embedded navigation, from the funding application to the accepted deliverables.
Skills
- Programming
- Go, Python, TypeScript (Deno), Kotlin (Android), Swift (iOS, macOS), Dart (Flutter), Ruby on Rails, Bash
- Platform
- Kubernetes (K8s, K3s, Talos), Helm, Argo CD with ApplicationSets, Terraform, Docker, Proxmox, NixOS, Linux
- Cloud
- Google Cloud for many years, AWS, basic Azure, hybrid on-prem setups
- Operations
- Traefik, cert-manager, Keycloak (OIDC/SSO), Harbor, HashiCorp Vault, Longhorn, S3 object storage, KEDA, Prometheus, Grafana
- Data and workflows
- PostgreSQL with CloudNativePG, Redis, Kafka, NATS, Temporal, event-driven systems, sagas, event sourcing
- AI
- private LLM inference, RAG, OCR and document extraction with confidence scores, evals against independent ground truth, MCP servers, tool-calling agents
- Finance
- bank APIs (PSD2), Stripe, PayPal, DATEV export, XRechnung, ZUGFeRD
- CI/CD
- Gitea Actions, GitLab CI, GitHub Actions, dependency and secret scanning
- Leadership
- engineering management, team building and hiring, QA organisation design, tech strategy, fractional CTO