Senior Product & Delivery Lead · Local-First AI Architect · Human-Orchestration Systems Frankfurt am Main · 15+ years of delivery across regulated fintech, mobility & public-sector

I operate at the intersection of complex regulatory frameworks and advanced local-first technology. My journey began writing Turbo Pascal on an Atari ST and learning English from a Game Boy and Dune II — then organizing early multiplayer communities that became a consequence-free training ground for leading distributed, cross-functional teams long before "remote work" was a corporate standard. (One of those games was Warhammer Online; I later worked at its European publisher, GOA Games — my name is in the credits somewhere, which still makes me grin.)

For over 15 years I've translated ambitious business visions into resilient, high-throughput execution across highly regulated environments — PayPal, Avira, Tipico, Velvon, and Cubic — from cloud-native core banking under BaFin scrutiny to public-transport platforms serving millions. The role stays constant: stand between ambitious stakeholders and complicated systems and make them cooperate, transparently.

Today my engineering focus is sovereign, local-first AI architecture. I design air-gapped, GDPR-compliant agentic substrates on Apple Silicon — deterministic-first ingestion and execution that guarantees data integrity, privacy, and performance without fragile cloud-dependent APIs. Everything runs on Mac-as-Code principles (chezmoi, versioned manifests) and local GitOps, so the infrastructure and the local LLM runtimes alike are fully reproducible, auditable, and resilient.

I lead hybrid orchestration systems where human teams and local AI agents collaborate under strict engineering guidelines. The crew is just different now — these days I summon part of it, and right now that work honestly feels a little like being a cabbalist, a conjurer of constructs. That's only the current name for it; the constant is the threshold between business and technology, and I've stood there my whole life.

I'm based in Frankfurt am Main, the town I was born in, raising my daughter in a world that's AI-native from the start — and trying to be an AI-native dad worth having. (My wife taught English once, in Brazil; now she's at Nintendo, building the machines that do the teaching.) I build open, sovereign, secure systems made to stand up to regulatory and architectural scrutiny — honest work, out in the open.

Most "human-in-the-loop" AI treats the human as a brake on an autonomous agent. I'm working on the inversion: the AI orchestrates the human. It holds the situational picture and does the framing; the human keeps the authority and acts. There's real evidence behind this — Fügener et al. (Information Systems Research, 2022) found human–AI teams beat the AI alone only when the AI delegated to the human, because people systematically misjudge their own competence.

Applied to delivery and people-leadership, that becomes a system that runs a manager through a living, rule-anchored playbook — grounding every claim against a verifiable source, anchoring every decision to a citable rule, and surfacing exactly where a human left something implicit. Three principles keep it trustworthy:

• Grounding over fluency — deterministic ground truth first; the generative layer is fallible and always scored against the source.
• Symmetric accountability — the same machinery that audits a stakeholder audits the operator too, on identical terms.
• Name dysfunction, don't launder it — when the conditions for honest work are absent, the system says so instead of producing a tidy plan that pretends otherwise.

I name my tools so I remember them — the name carries the job. The Mechanicus flavor is deliberate: a dense, self-consistent vocabulary compresses well, makes rules behave like dogma a model follows reliably, and turns any out-of-canon term into a cheap tripwire for tampering. Where lore and engineering disagree, engineering wins. (All 40k marks © Games Workshop; unofficial, non-commercial.)

The internal Git origin is the authority; this profile is the showcase. Private systems are described at the architecture level only.

• Sovereign synthesis engine. A deterministic-first pipeline that turns heterogeneous source documents into grounded, rule-anchored backlogs — and, increasingly, into a living playbook that orchestrates the human running the project. Demonstrated: a hand-written 36-story backlog reproduced as 29 fully rule-traceable stories at a 0.00 hallucination score, catching errors the human-written original contained.
• GitOps control plane. Idempotent bare-metal hardening, network partitioning, isolated container stacks, sops+age-encrypted secrets in-repo (the key never leaves the control plane), and a sovereign private origin. If the substrate died tomorrow, it rebuilds from git to a known-good state.
• Contactless transit-payment architecture (client-confidential). Open-loop fare-payment integration aligned with Visa Contactless Transit (MTT) and Mastercard Transit guidelines, delivered as a version-controlled feature backlog and technical specification set.

A federated, two-node environment I control end-to-end, so private data stays on hardware I own and every AI output traces back to a deterministic ground truth.

• Control plane — Apple Silicon. Agentic orchestration, local LLM inference (Ollama), deterministic document synthesis. Reproducible via chezmoi + a versioned Brewfile.
• Substrate — bare-metal Ubuntu, deliberately disposable. Isolated container stacks, hardened host, private Git origin with FTS5 retrieval.
• The doctrine — code moves by GitOps, never by file-sync. The control plane is the truth; the substrate is reconstructable.

• Led the turnaround of a flagship public-transport app from 1.6 to ~4.0 stars, scaling to 3M+ users through the politically critical Deutschland-Ticket launch with zero downtime — and secured €2M+ in annual renewals by aligning delivery to the client's fiscal reality.
• Delivered audit-ready, regulator-grade software for a cloud-native core banking system on GCP, mapping every component to its regulatory requirement (BaFin / KWG) to clear pre-license audits — proving a German banking core could run compliantly on a US hyperscaler.
• 15+ years across fintech, public sector, cybersecurity, and mobility. PRINCE2, PSPO, PSM I; Executive MBA. Bilingual in German regulatory rigour and US/SaaS commercial agility.

I'm most useful where high stakes, heavy regulation, and powerful new technology meet — and where someone has to make all three coexist honestly. I'm pointed at public-sector / B2G and EU-institution leadership and principal/advisory roles — Berlin, Brussels, Frankfurt — turning digital sovereignty, AI governance, and privacy-by-design from buzzwords into architecture that actually holds. The orchestrator who stands between leadership, regulation, and engineering and makes the three cooperate.

• LinkedIn — linkedin.com/in/mauricejobst (best place to start a conversation)
• GitHub — github.com/maurice-jobst
• Email — maurice.jobst@gmail.com

Built in Germany, on silicon I own. Honest work, out in the open.