Recarta: Intelligent Institutional Real Estate Management

The startup Recarta has quickly established itself as a specialist in digital data extraction and processing for institutional real estate management. The Swiss company focuses on an economically significant niche: institutional real estate managers. These clients—including pension funds, publicly traded real estate funds, cantons, and international law firms—are grappling with a fundamental problem: How can they transform the unstructured data from the overwhelming volume of documents related to their real estate portfolios into actionable information?

The Search for a Modern, Robust Solution

Recarta recognized that the biggest problem was not the volume of data, but its unstructured form. Better Excel spreadsheets or traditional management software would not solve the problem—a fundamentally different technology was required. Furthermore, Recarta was guided by a fundamental requirement in its search for a solution: digital sovereignty. For this reason, the solution could neither be based on U.S. cloud infrastructures nor result in dependence on a specific provider (“vendor lock-in”). The former is important due to the sensitivity of the data being processed, while the latter offers the company greater flexibility and future-proofing.

Strategic Partnership with Exoscale

Given these requirements, Recarta sought a European partner with the appropriate technological expertise and found it in Exoscale, the European cloud provider. Together with the Swiss Data Science Center, the University of Applied Sciences and Arts of the Canton of Vaud, and Exoscale, Recarta systematically evaluated open-source models such as Mistral, Qwen, and Llama under realistic conditions in multiple languages. The result: Open-source models outperformed proprietary systems on Exoscale’s infrastructure, proving that sovereignty and technological excellence are not mutually exclusive.

Core Technical Solution

The jointly developed solution converts unstructured documents into structured data. Using AI, the system automatically extracts 20–30 relevant data points from a 20-page lease agreement: signatories, date, price, renewal clauses, etc. From cost estimates and invoices, it captures all items, quantities, and prices. From permits, it extracts critical parameters such as location, type of use, and regulatory requirements.

The solution has also overcome a technical challenge that arises when using standard LLMs: the lack of determinism—where the same inputs lead to different results. The partners overcame this challenge through specialized prompt engineering, intelligent orchestration, and validation mechanisms on Exoscale’s dedicated infrastructure. The result: the system delivers the same result every time for the same input.

Dedicated Inference as a Strategic Advantage

Exoscale provides Recarta with Dedicated Inference Services for operating its solution. This allows Recarta to run its own LLM instances of choice on dedicated Exoscale hardware without having to share resources, as is the case with OpenAI, Azure, and AWS. Specifically, Dedicated Inference avoids the so-called “noisy neighbor” effect: With shared resources, an application’s performance can fluctuate if another customer is placing an extreme load on the same hardware. With Exoscale, the booked GPU belongs solely to the respective customer, resulting in extremely stable and predictable response times. Additionally, the physical or logical separation of data streams with Dedicated Inference ensures greater security and better compliance. Because the hardware does not process requests from other customers simultaneously, the risk of side-channel attacks or data leaks between tenants is technically minimized. Last but not least, Recarta has full control over the instance: The company receives a dedicated GPU instance that operates exclusively for its private API endpoint. It can specify exactly which model is loaded and how resources are used, without the provider’s load balancer queuing its requests alongside those of thousands of other users.

Versatile Use Cases

The solution covers multiple scenarios. In contract management, the system automatically extracts notice periods and renewal conditions and generates alerts for critical deadlines. In due diligence processes for real estate purchases, it screens hundreds of documents for risks such as historic preservation requirements, environmental contamination, asbestos exposure, or liability obligations. In cost management, invoices are automatically compared with cost estimates, and discrepancies are identified. In the compliance area, the system comprehensively checks for adherence to guidelines. The Canton of Vaud, for example, already uses Recarta to perform consistency checks on building permits and automatically identifies contradictions between documents.

Significant Advantages

The Recarta solution saves a tremendous amount of time. Document analyses that used to take hours are now completed in minutes. In addition, automatic consistency checks eliminate human error and ensure accurate compliance documentation, while automatic contract monitoring prevents missed deadlines or the need to adjust expiring contracts. This can result in annual savings in the five-figure range. Furthermore, decision-makers can now view all key information via centralized dashboards and no longer have to spend time searching for data scattered across various locations. This allows them to perform comparative analyses between cost estimates and bids: The system highlights the differences between the original budget and the various bids, calculates averages, and generates automatic alerts—for example, if a bid contains additional items that were not included in the original cost estimate.

Conclusion

With high-profile clients, Recarta and Exoscale are demonstrating that European technology ecosystems can be innovative while also reconciling data sovereignty and independence. The project serves as a model: Recarta identified a specific pain point in property management and, together with Exoscale, the Swiss Data Science Center, and the University of Applied Sciences and Arts of the Canton of Vaud, developed an elegant solution based on European infrastructure. This also shows other European startups a way forward: genuine technological innovation based on European sovereignty is not only possible but can also deliver better performance than the supposed market leaders from the U.S.