Zeit AI raises €5M to bring an autonomous data engineer to Europe's mid-market

Why it matters

The round tests whether Palantir-style forward-deployed engineering can be compressed into software for midsize companies without leaving Zeit AI trapped in services-heavy deployments.

Leopold von Waldthausen (@lwaldthausen) and Marvin Christopher Bornstein, former Palantir colleagues who spent years deploying software inside large European businesses, have raised €5 million for Zeit AI. The Munich startup is trying to package that expensive, labor-intensive work into an AI data engineer that midsize companies can buy.

The financing, reported on September 3rd by Tech.eu, includes Y Combinator, the University of Oxford's Seed Fund, Sequoia Capital Scout Fund, ACE Ventures and Hasso Plattner VC. Angels in the round include former German finance minister Christian Lindner, footballer Mario Gotze, Charlie Songhurst (@charlie), Martin Schoeller and Helsing CTO Robert Fink.

Zeit AI plans to hire engineers and expand deployments of ZeitMind, which connects data held across ERP, CRM, HR and other operational systems. Zeit AI says the product supports 600+ sources, cleans and links their data, and builds dashboards, reports and operational applications from instructions written in natural language.

Von Waldthausen and Bornstein founded Zeit AI in 2024 after working side by side at Palantir. Von Waldthausen spent 3.5 years there and eventually led the established customer business across Germany, Austria and Switzerland. He had already founded a video-review startup, weview, at 18 and sold it to DemoUp Cliplister after three years. His path also included computer science at Oxford and an MBA program at Yale that he left before completion.

Bornstein spent five years at Palantir as a forward-deployed engineer and technical lead. He says he deployed early versions of the technology that became Palantir's Artificial Intelligence Platform across industries including chemicals, rail infrastructure, utilities, insurance and medical equipment. Before Palantir, he studied IT systems engineering at the Hasso Plattner Institute and worked in software development and deep-learning research.

Zeit AI says projects the founders worked on at Palantir produced €50 million in new revenue. Zeit AI has not published enough detail to independently allocate that revenue to particular deployments, but the experience explains the founders' product thesis: valuable enterprise data work still depends on engineers entering a customer, understanding its processes and connecting systems that were never designed to cooperate.

A Palantir deployment model for smaller customers

The founders left Palantir in summer 2024 after repeatedly seeing finance, procurement and plant teams move data between enterprise systems and spreadsheets by hand. Large corporations can assign data engineers and consultants to those problems. A midsize manufacturer or logistics operator may have the same fragmented SAP and CRM records without the budget or recruiting pull to assemble a specialist data organization.

ZeitMind is aimed at that gap. Zeit AI says a customer can describe a requirement such as combining SAP financial records with CRM pipeline data, and the agent will connect the sources, prepare the information and build an application around it. The resulting figures are designed to remain traceable to individual transactions. Zeit AI says customer data is hosted in Germany and is not used to train models.

That scope separates ZeitMind from agents focused on one layer of an existing data stack. TensorStax, for example, documents agents that create and edit dbt models, Airflow workflows and AWS Glue scripts, with restrictions on destructive database operations. Definity, which announced a $12 million Series A in April, focuses on operating and optimizing Spark and lakehouse pipelines during execution.

Zeit AI is making a broader bet. ZeitMind is supposed to move from raw operational systems through data preparation and into the business application used by a controller, buyer or supply-chain manager. Owning that whole path gives Zeit AI a larger contract opportunity and a larger failure surface. Connecting a source is useful; producing a number that a finance chief will defend in a board meeting requires lineage, permissions and consistent behavior as the underlying records change.

Autonomy still comes with engineers

Zeit AI's commercial model also reveals what "autonomous" means in practice. Customers are being sold freedom from hiring their own data-engineering headcount. Zeit AI still supplies human deployment work.

Its website says a forward-deployed engineer puts ZeitMind into production with finance, procurement and supply-chain users. The newly raised capital will fund additional engineers who can support those deployments. Zeit AI is effectively centralizing scarce engineering talent inside its own organization, then using the agent to let each engineer handle substantially more customer work.

That is a more credible near-term proposition than handing an unfamiliar agent the keys to a live ERP system. It also means Zeit AI must prove that deployments become less labor-intensive as the product matures. Otherwise, revenue growth will continue to require a matching expansion in expensive technical staff.

Bornstein has published unusually specific detail about how ZeitMind attempts longer tasks safely. In an August 10th technical post, he described a branching system that lets the agent make a chain of changes across tables, pipelines, charts and applications without writing directly into production. A typical run can span 100 to 300 iterations using 45 tools, followed by a single review of the resulting tables and visualizations.

The post also documents the unfinished work. ZeitMind did not yet detect conflicts when production data and an agent branch changed the same table, abandoned branches required manual cleanup, and Zeit AI was still developing rules governing who could merge changes. Those details do not invalidate the automation claim. They define the engineering required before an agent can safely replace repeated human approvals in systems carrying financial and operational data.

What the €5M buys

Zeit AI joined Y Combinator's Summer 2024 batch. Its current YC profile lists 12 people across Munich and London. Zeit AI says it reached €1 million in annual recurring revenue within 12 months with six employees, a self-reported figure that does not include a customer count, pricing breakdown or current run rate.

The investor group reflects both founders' routes into Zeit AI. Y Combinator backed the initial company-building sprint. Oxford's fund follows von Waldthausen's university connection, while Hasso Plattner capital matches Bornstein's academic and technical background. Sequoia's scout network and operators including Songhurst and Fink add relationships beyond Germany's midsize industrial base.

The round gives von Waldthausen and Bornstein room to recruit the engineers needed to deploy ZeitMind while they turn repeated customer work into reusable product capability. Their central challenge is economic rather than cosmetic: each new deployment must teach ZeitMind enough that the next customer requires less custom engineering.

Zeit AI has chosen a market where the pain is durable and the systems are messy. Spreadsheets persist because they are flexible, visible and controlled by the people doing the work. ZeitMind will earn its place by matching that flexibility while preserving the traceability and safeguards required when an AI agent starts changing the data products behind real operating decisions.