Tech Lead / CTO (m/w/d)

The opportunityTax advisory has everything automation needs: structured data, recurring work, clear rules. Yet the profession never made the leap, because recurring does not mean uniform – every case has its own twist, and the law keeps moving. Reasoning resolves exactly this variance. That is why the window is open now. At the same time the market is being reordered: four in ten tax advisors in Germany leave the profession within ten years, and many of their firms will find no successor. Vogel acquires established advisory firms and develops them further. Several thousand mid-sized companies trust us with their tax, accounting and strategic matters, and every acquisition adds to that base. We have already produced our own annual financial statements with AI. Today, that is one person with a tool. Turning it into a capability of the whole company is the job. You own the technology that makes Vogel AI-first: the data layer above DATEV, the agent infrastructure on top, and the control layer without which nothing in this profession may go live. You decide the architecture and the stack of what we build. You report to and work directly with Moritz Vogel. Decisions are made in minutes, not months. What you'll doArchitecture above DATEV Build the data layer above DATEV. DATEV stays the system of record (for now); the value is created in the layer above it – designed to outlive the system beneath. Choose and build the agent infrastructure: framework, orchestration, versioning, observability. You answer these questions with code, not with a slide deck. Make integration a capability, not a project: every client brings their own upstream systems, and today each connection is built by hand. Build the layer those systems plug into, so what makes a client different arrives as context the agents read – never as custom code. One more client must never mean one more integration project. Reproducibility and control A tax advisor must get the same result twice, and an auditor must be able to trace it. Probabilistic where there is variance, deterministic where there are numbers. Build the control layer so that a certified professional sees within seconds what an agent did – version, source, decision – and can sign off on it. Build evaluation on what no newcomer has: decades of audited, completed work. Team and scale Make the architecture calls: design, stack, build or buy for the layer. Nobody has decided it for you, and nobody will overrule it by committee. Hire and lead the engineering team. The budget for your first engineers is approved – you start hiring in your first quarter. Make every acquired firm land on the same base, so technology scales with every acquisition. And stay in the code: you commit and you review. Open questions – yours to answerWhere exactly does probabilistic end and deterministic begin? How does a system learn from closed cases without losing its reproducibility? A new client arrives with their own systems. What may change in the system, and what may not? What success looks like6 months: Architecture decided. One real use case from the core business runs in production and is used daily. Your first engineers are hired. 12 months: Several agentic use cases in regular operation. The control layer holds up in a real review, and evaluation runs against closed cases, not test data. 24 months: Automation is a capability of the organization, not of individuals. Every acquired firm runs on the same base, and your team ships without you in the room. Where it leads: the full CTO mandate – a management career earned by contribution, not by years served. What you bringYou have built something that ran in production and was used by people who did not build it – as founding engineer, engineering lead, technical co-founder or in a comparable role. You have taken agentic systems into production and know where they break. You have integrated against systems that fight back: legacy ERPs, banking backends, badly do