The technical architecture of PropTech

The German commercial real estate market remains characterized by manual, paper-based processes and fragmented data storage. domos, founded in late 2023 and based in Berlin, enters this market with a specific focus on translating this analog history into structured digital assets. The company is led by CTO Lukas Mateffy, a software architect whose background in PHP development and robotics informs the platform’s engineering-heavy approach to property management.

At the core of the domos platform is a detailed data model designed to store more than just basic property listings. It is built to support a comprehensive understanding of real estate assets for both human agents and machine learning models. By digitizing documents like lease agreements, floor plans, and technical certificates, domos enables real estate professionals to automate the creation of marketing materials and the analysis of property performance. This automation is not a skin-deep feature; it is built on a custom SDK developed by Mateffy called LLM Magic.

Solving the PHP-LLM gap

While much of the current AI agent ecosystem is built on Python or TypeScript, domos is notable for its commitment to the Laravel and PHP ecosystem. Mateffy’s creation of LLM Magic stemmed from a practical deficiency: existing OpenAI libraries for PHP were often restricted to simple chat interfaces and lacked native support for advanced features like tool calling, streaming, and model-agnostic switching.

LLM Magic provides the underlying engine for the company’s "Data Wizard" functionality. It uses reflection and PHP-specific features to automatically generate code for tool calling, making it easier to swap providers like Anthropic, OpenAI, or Google Gemini with a single line of code. This flexibility is critical for a real estate platform that must process documents of varying quality and format, where one model might excel at OCR while another is better at logical reasoning across a multi-page PDF.

Extracting structure from chaos

The primary value proposition for domos users is the ability to turn unstructured data—PDFs, images, and scanned documents—into validated JSON data. Users provide a schema, and the platform handles the chunking, context management, and extraction strategies required to fill that schema accurately. This process removes the need for manual data entry, which is the traditional bottleneck in commercial real estate management.

The platform's target users are commercial agents and landlords who manage large portfolios. Instead of "yelling at an LLM" to return properly formatted data, domos handles the validation and UI generation automatically. This pragmatic application of agents—treating them as reliable data-entry and processing layers rather than general-purpose assistants—marks a shift toward the specific, vertical AI applications necessary to modernize legacy industries.