Architech-AI
Where the Art of Design Meets the Science of AI.
Created on 24th January 2026
•
Architech-AI
Where the Art of Design Meets the Science of AI.
The problem Architech-AI solves
- "Design, to, Data" Gap
Usually, an architect makes a design and an engineer a few days later says it is not stable. Architech, AI solves this problem by giving structural feedback instantly. The calculation of slenderness ratio and wind load is done while you move the slider, thus impossible designs are stopped even before they reach the drafting table.
- Logistics and Cost Guesswork
Estimating the quantity of materials (cement, steel, bricks) is one of the things that usually require very intricate spreadsheets. This project automates Material Takeoffs (QTO), thus giving a user an instant "Bill of Quantities" based on the 3D geometry being looked at.
- Environmental Misalignment
Architects are sometimes at a loss when trying to visualize the interaction between sunlight and the mass of a building. The Solar Study feature helps designers get the most out of natural light and heat gain by performing shadow analysis in real, time without having to resort to expensive simulation software.
- Loss of Artistic Intent
Contemporary CAD programs can appear "uncaring" and entirely the product of mathematics. The project stays close to the artistic inspiration through charcoal and graphite hand, drawn sketches incorporated in the Pencil Reference Gallery while at the same time allowing for high, tech digital modeling.
Challenges I ran into
- 3D Mathematical Performance
Challenge: Rendering multiple floors with very realistic glass refraction (transmission) uses a lot of GPU power. Workaround: We have tried to work buffer usage with React Three Fiber and Dreis MeshTransmissionMaterial to a point where the "Heatmap" and "Glass" modes can switch instantly without any browser lag.
- Bridging Different Environments
Challenge: How to make a Python, based engineering engine communicate with a JavaScript, based visualization engine in real, time. Workaround: FastAPI REST architecture was implemented with quite optimized JSON payloads. Pydantic models were used to guarantee the complex architectural data types understood between backend logic and frontend 3D rendering are the same.
- Realistic Shadow Mapping
Challenge: Shadows created by WebGL only look "jagged" or "pixelated" most of the time, resulting in an architectural "look" being ruined. Workaround: We combined Soft Shadows and a Dynamic Sky System. We narrowed the shadow camera's viewing volume just around the building so that we got very clear shadows even with the "Sun" at a low angle.
- UI/UX for "Data Overload"
Challenge: Putting the information such as wind load, cost, efficiency, materials, and a 3D view all together on one screen is too much for the average user. Workaround: We went for a Dashboard, First Layout. By