AI-based editor for accurate architectural specifications
Services:
Project overview
Architects work with different complex specifications that require frequent updates and revisions, in particular:
- Building codes and regulations
- Construction methods and techniques
- Client and stakeholder requirements
- Contract and legal documentation
- Environmental and sustainability standards
- Zoning laws, land use, site analysis, fire safety, and many other requirements
The curation and management of all these documents is cost- and time-consuming, as well as prone to errors. To resolve this issue, we designed a smart, web-based platform to simplify the process of assembly and editing by leveraging large language models (LLMs).
Architects upload their specifications, select sections to modify, and provide further instructions and remarks. The solution then analyzes the content, suggests adjustments, and allows for iterative section fine-tuning.
Once perfected and completed, architects can quickly download the updated, easy-to-manage specification. This tool can facilitate the revision while minimizing human error.
Main goals
By integrating large language models (LLMs), we empower architects dealing with specifications by allowing:
- Faster revision
- More accuracy
- Greater focus on design and creativity
- Regulatory compliance
By automating and optimizing the process of editing, we help architects struggling with their daily documents to boost everyday efficiency and elevate the quality of projects.
How the solution works
By providing these features:
- Document upload and mark-up, notes understanding
- Section editing
- LLM-based conversational interface
- Smart revision suggestions
- Real-time document updates
- Finalized specification download
We offer a straightforward, efficient process to manage architectural specifications:
1. The user first uploads the document in a PDF/DOCS format
2. Once uploaded, the document is processed for interaction
3. The architect can select specific sections to modify and focus on the important areas
4. After selecting the section, the architect then engages with the AI editor
5. By using retrieval‑augmented generation and advanced prompt engineering, our solution will provide intelligent suggestions, answer questions, and assist in refining provided specifications
6. Once completed, the document can be quickly downloaded with already incorporated changes
Our contribution
Our project was aimed at providing architects with a solution to boost their work by leveraging AI assistance. We participated to design and implement the components already described, to ensure seamless integration and smooth, user-friendly experience.
We started with a narrow focus on resolving editor notes and designed an editor to cover broader application.
We covered:
- The development of the core functionality, which allows to upload, easily edit, and download architectural specifications
- LLM implementation (OpenAI, Claude) to enable real-time editing
- The design of a custom system, which handles document handling and processing
- The management of both data analysis and storage to maintain document security and integrity
- And ensured compliance mechanisms are all in place, thereby allowing the platform to align the output with industry-specific regulatory standards
Main challenges
AI integration
LLM integration came with complex hurdles with real-time data processing and ensuring smooth interactions. We optimized the pipelines to enhance overall responsiveness, thereby ensuring seamless communication between models and workflows.
Prompt engineering
Prompt engineering was essential to generate adequate suggestions complying with architectural standards. We developed both specialized prompt structures and iterative testing methods to refine provided responses for accuracy and alignment.
User experience & usability
Another problem – to conceptualize and build an editor to upload, easily edit, and download the specifications. We designed a simple, user-friendly interface with easy, intuitive navigation, clear functionality, and real-time editing capabilities.
Document formatting & export
And finally – the retention of modifications after exportation, which required careful attention to compatibility. We implemented advanced techniques to preserve document structure and formatting, and compatibility across different file types.
Tools & technologies:
- Python
- OpenAI API
- Claude API
Timeline:
- June 2024 – August 2024
Team:
- 1 project manager
- 1 AI engineer
Value delivered to business
Our platform is set apart from similar software by several unique benefits:
- Conversational editing – unlike other traditional solutions, our platform allows architects to interact with the AI solution in real-time, refining specifications through natural language input
- Revision suggestions – with this custom platform, architects access context-aware recommendations aligned with architectural standards
- Regulatory compliance – the system sensibly reduces legal risks without requiring manual verification
- Preserved formatting & structure across different file types during editing and export