// case 03 · AI Engineering · 2025
BIM Forum El Salvador
RAG conversational assistant that transforms natural language queries about BIM technical standards into precise answers with citations and page numbers — reducing from 20–40 minutes to seconds. Publicly presented at BIM Forum El Salvador 2025.
// The challenge
Architects, engineers, and construction professionals spend 20–40 minutes navigating dense, multi-volume technical normative documentation every time they need to verify a BIM standard. This manual search exposes professionals to interpretation errors with costly consequences on construction projects. BIM Forum El Salvador wanted to demonstrate AI's potential in the sector — not as an experiment, but as a real product tool with engineering criterion.
// what we built
Semantic embeddings over complete technical standards, persistent vector store with ChromaDB, and LLM integrated with engineering criterion. Not a proof of concept — a robust, maintainable system.
Every response includes the exact source and page number, enabling immediate verification. This generates direct traceability over design decisions and eliminates the risk of incorrect interpretations.
Conversation history, dark theme, responsive design, and session persistence. An interface reflecting deliberate product decisions, not an improvised demo.
Independent services with centralized configuration, request validation, and documented REST endpoints — ready to integrate with future sector systems.
The project was publicly showcased at the BIM Forum El Salvador 2025 conference, demonstrating generative AI's potential in construction to an audience of sector professionals and industry leaders.
// The impact
Time to query technical standards
Responses with verifiable citation and page
Presented at BIM Forum El Salvador
BIM Assistant turns a daily and costly construction-sector problem into an operational advantage. A query that previously took 20–40 minutes of manual searching resolves in seconds, with a cited, verifiable answer. This reduces the risk of normative interpretation errors, democratizes access to technical knowledge within any team, and generates direct traceability over every design decision — what was previously just good intention is now automatic.
// are you next?
A 30-minute call, no deck. We'll tell you honestly whether we can help you and how.