Selected Work

I have spent sixteen years building software across startup zero-to-one mandates and high-traffic enterprise platforms. The common thread across those roles has been owning the boundary where system reliability, data modeling, and developer leverage meet.
Production AI Platforms & Systems
European Real Estate AI Platform (2026)
Role: Principal AI & Systems Architect (Independent Mandate)
Scope: Document Cloud Workflows · Multi-Agent Governance · Deterministic Verification Gates
Architected core runtime and quality-gate systems for a European enterprise AI platform processing regulated real estate and diligence documents at scale.
- Designed structured agent-orchestration harnesses combining schema validation, deterministic verification gates, and audit trails across model providers.
- Re-architected multi-stage document cloud ingestion and asynchronous processing pipelines for predictable latency and observable failure recovery.
- Built repeatable regression suites and release discipline for mixed human-plus-agent engineering delivery.
agent-fabric & command-center (2026)
Role: Creator & Maintainer · Open Infrastructure
Stack: Python · TypeScript · SQLite · Git Worktrees · Multi-Provider CLI Orchestration
Built an open coordination substrate for running multiple coding and verification agents concurrently without shared-state corruption:
- Lease & worktree governance: SQLite-backed task claims, per-worktree path leases, and clean integration boundaries across isolated agent checkouts.
- Independent review gates: Automated cross-provider verification policies bound to plan digests and exact base SHAs.
- Operator telemetry (
command-center): Unified cockpit monitoring agent runs, cost/quota usage, and gate results in real time.
Samsa.AI (2024-2025)
Role: CTO & Lead GenAI Architect · Zurich (Remote)
Designed and shipped the production AI/ML platform combining multi-provider inference routing (fal, Replicate, Hugging Face), custom PyTorch GPU workers, and multi-agent generation pipelines across AWS, Cloudflare, and Supabase.
Founding & High-Scale Track Record
auteon (2019-2020)
Role: Founding CTO · Software Architect · Munich
Took an automotive B2B marketplace from an empty editor to production deployment in six months: designed the proprietary automotive parts data schema, scaled to 400+ active industry users, and led technical due diligence that closed seven-figure seed financing.
PeterPark (2020-2021)
Role: Lead Engineer · SRE · Advisor · Munich
Architected cloud infrastructure on AWS for IoT hardware-integrated parking systems, led engineering delivery, established shared UI foundations, and reduced heavy analytical PostgreSQL query latency from ~15s to sub-500ms.
Sixt & Sixt ride (2015-2017)
Role: Senior Front-End Manager & Technical Lead · Munich
Led the frontend architecture reboot moving booking experiences off legacy PHP templates onto a modern React SPA serving 1M+ monthly bookings, and directed a microservices API consolidation reducing redundant endpoints across mobility services.
Additional Mandates
- GrowMetrics (2023-2024) & StudySmarter (2022-2023): Interim CTO / Enterprise Architecture for cloud cost optimization, CI/CD governance, and high-scale EdTech database/service decomposition.
Open-Source Developer & AI Engineering Tools
I build sharp developer tools to enforce determinism around probabilistic systems:
| Project | Purpose | Documentation & Source |
|---|---|---|
| AntiSlop | Tree-sitter AST linter across 15+ languages for AI-generated placeholders & stubs | GitHub · Docs |
| EnvCheck | Fast Rust .env validator syncing with Kubernetes, Terraform, Helm, and CI | GitHub |
| NerfStatus | 17-probe capability regression & silent collapse detector for hosted LLMs | GitHub · Dashboard |
| HCT Signals | Polyglot coordination protocol extensions for MCP & A2A (Python, TS, Rust, Go) | HCT Spec · Research |
| SpeakOps | Low-latency hybrid local/cloud speech telemetry & coaching assistant | GitHub |
Explore full Tools documentation → · Explore Open Source index → · About & Contact →