My Projects

My Project space

Canvas M: Meeting Intelligence for AEC Teams

AI is only as valuable as the data it learns from. Rent someone else's intelligence, trained on someone else's data, and generic answers are all it can give back. An AI that truly understands your projects needs your own data first: centralized, structured, and owned. Data is the asset. In AEC, that asset is born in meetings. Decisions, reasoning, revisions, agreements.

But the way teams meet today scatters all of it the moment it is created, across call tools, chat threads, drawings and emails. For a practice with 12 branches working over many time zones and languages, the scatter multiplies. What a project decided last quarter often survives only in one person's memory. Canvas M grew out of that question at MAPGROUP and has just been completed as a milestone project at our Research Center for AI in Architecture. Meeting data is captured at the source, remembered, and linked across meetings, so over time it becomes the base for a company's own AI.

What it does:

▸ Everything an online meeting tool needs (video, chat, scheduling) plus a shared infinite canvas

▸ BIM (IFC), CAD, PDF, image, audio and video: the working files of a construction project, on one canvas

▸ Project-based meeting organization and participant management ▸ Real-time transcription and translation, so everyone reads in their own language

▸ AI inside the canvas that understands the meeting and answers questions about it, with sources

▸ Full automatic meeting recap

▸ Project Knowledge: every meeting adds to a project-wide memory, so the AI understands the meeting, the project, and starts proposing what comes next And because that memory keeps growing, one day someone will ask in the middle of work: "That façade detail on the project four years ago, how did the team agree to resolve it?" The answer comes back in seconds, with the original meeting as the source. The kind of experience that used to leave with people now stays with the project. Shaping it from concept to design to code has been the most rewarding stretch of my work so far.

Your data. Your knowledge. Meetings end. With the right system, their memory doesn't.

Physical-Digital Integration Framework | Real-Time Wind Flow Prediction

What if you could design with physical models and instantly see how wind flows around your buildings? This breakthrough framework bridges the gap between hands-on design and real-time environmental analysis.

Watch how our breakthrough framework enables architects to design with physical building blocks and instantly see wind flow patterns using deep learning.

🚀 What You'll See:

✅ Real-time wind flow prediction in below 1 second

✅ Physical model manipulation with 3D-printed apartment blocks

✅ AI-powered multimodal pix2pix architecture

✅ Natural language parameter control ("I wonder how the wind would affect this design in Seoul in winter.")

✅ ArUco marker-based spatial recognition

📖 Read full paper: "A Physical-Digital Integration Framework for Environmental Simulation through Deep Learning: Wind Flow Implementation"

DOI:10.1016/j.buildenv.2025.112869

👨‍🔬 *Lead Researcher:* Thanh-Luan Le

🏛️ *Institution:* Sungkyunkwan University (South Korea)

📚 *Google Scholar: https://scholar.google.com/citations?user=nfQu98IAAAAJ&hl=en

📬 *Contact:* lethanhluan@skku.edu

Automated Factory Planning Platform

A joint research initiative between MAP and DIG (SKKU) aimed at developing a web-based automated factory layout optimization platform. The system integrates optimization algorithms, automation workflows, and browser-based 3D visualization using Three.js, supporting efficient decision-making in early-stage factory planning.

The system transforms early-stage factory design inputs, including datasheets and site plans, into optimized layouts, parametric 3D models, and AI-enhanced renderings, significantly reducing planning time while improving design quality.

👨‍🔬 Lead Researcher: Thanh Luan, Le

📚 Google Scholar: https://scholar.google.com/citations?user=nfQu98IAAAAJ&hl=en

📬 Contact: arch.leluan@gmail.com