Automatic generation of architectural projects

The automatic generation of architectural projects is at the heart of Davis’s strategy, which announced a €4.6 million pre-seed and the first version of its Gaudi-1 model.

Why Davis is betting on Automatic Generation of Architectural Projects

In the real estate sector, initial workflows often remain slow and fragmented, and Davis positions itself as a solution to accelerate them by integrating proprietary AI and human know-how.

The operating promise of Davis is to compress conceptual design times from weeks or months to a few hours or days, while maintaining human-quality control.

Financing and partners

The €4.6 million round was led by Heartcore Capital and Balderton Capital, with investors including Yellow, Evantic and Entrepreneurs First, plus angels from SpaceMaker, Hugging Face and others.

International backing and the presence of tech angels signal strategic confidence in the AI-and-real-estate combination proposed by Davis.

Technology: a discrete model and Gaudi-1

At the center of the proposal is Gaudi-1, a model that generates buildings not as continuous pixels but as discrete sets of architectural elements — rooms, walls, layouts and volumes.

Generating buildings in a discrete space allows greater control over design constraints and faster iterations compared to traditional diffusion models in pixel space.

Technical advantages of the discrete paradigm

Davis states that the discrete model yields outputs that better comply with regulatory, financial, and design requirements, offering measurable metrics on IoU, FID and KID in floor-plan benchmarks such as RPLAN and MSD.

The practical result is floor-plan generation with competitive benchmark performance, translating into fewer redesigns and lower design risks for developers.

Gaudi-1 is not limited to creating images: it generates structured representations that facilitate regulatory verification and the estimation of economic return.

How Davis integrates the model into the real process

The startup does not sell a simple software: it offers a service that delivers outputs ready for developers and investors, with a human validation step before delivery.

The model is used as an internal engine within a service that produces feasibility studies, volumetrics and floor plans verified by architects before final delivery.

From data to constraint: concrete inputs

Davis transforms regulatory data, technical restrictions, and market parameters into structured constraints that guide the automatic generation of architectural projects.

Translating rules and financial parameters into constraints enables project proposals that are immediately assessable in terms of ROI and regulatory compliance.

This conversion of rules into constraints enables rapid iterations and scenario comparisons, reducing uncertainty in early real estate development stages.

Business model and go-to-market

Davis adopts an added-value service aimed at developers and investors, not just a software license: outputs are delivered complete, with local regulatory adaptation.

The commercial strategy focuses on selling outcomes and speed, not just licenses, making adoption easier for those who must make rapid investment decisions.

Partnerships and initial use cases

The startup is collaborating with developers on real-world projects and plans to support hundreds of projects in the coming months, testing Gaudi-1 across different asset classes and jurisdictions.

Testing Gaudi-1 on real-world cases helps refine the models and demonstrate time and cost savings before broader commercial scale.

Practical implications for founders and innovators

For those developing proptech solutions, the Davis case shows the value of modeling complex domains (regulatory, technical, market) as structured constraints that can be integrated with generative models.

Integrating a human-in-the-loop verification with a specialized generative model reduces risk and increases stakeholder confidence in AI-produced results.

Critical analysis: pros, cons and risks

Davis’s proposition is ambitious and presents clear benefits, but it also entails criticisms that need to be analyzed before investing or integrating similar solutions.

Below we discuss advantages and limits to provide a balanced assessment useful to founders, investors and technical teams.

Pros: the discrete model improves control and reliability of generated solutions; the service reduces decision times and facilitates risk quantification.

This approach can translate into fewer design revisions, faster early stages and better investment decisions.

Cons: dependence on local data and regulatory accuracy; generalization across different markets requires substantial constraint engineering and validation.

Rapid generalization without adequate regulatory localization can cause costly errors in permitting or economic estimates.

Technological and market risks: positive benchmarks do not guarantee real-world results; additionally competition for advanced generative models and protecting proprietary value are significant challenges.

To mitigate risk, Davis should maintain strong human validation and create repositories of replicable use cases for different markets.

In short, automatic generation of architectural projects has the potential to transform the early stages of real estate development, but requires deep integration of data, regulations, and design know-how to be truly effective.

Anyone considering similar solutions should weigh the trade-off between automation and control, invest in data localization, and maintain a robust human validation process.

Next steps and road map

Davis plans to expand research, hire talent, and verticalize the real estate process to cover more asset classes and jurisdictions in the next 12 months.

Operational priorities include scaling regulatory datasets, refining Gaudi-1 for different contexts, and solidifying partnerships with pilot developers and investors.

Market impact and operational conclusion

The emergence of models like Gaudi-1 signals a trend: specialized generative technologies are moving from demonstrations to products that solve concrete problems in real estate.

For founders and investors, the practical lesson is to focus on solutions that combine specialized AI, human validation, and a results-oriented distribution model, not merely the product.

Automatic generation of architectural projects represents today a concrete opportunity to reduce timelines and increase decision-making certainty in real estate development.

Source eu-startups.com