AI and Architecture | Research & Development
Type
Year
Professional (ArchiTAG)
Jul 2025 - present
ARCHITECTURE
AI
Architecture | Spatial Design | Artificial Intelligence
Overview
As a Research Associate at ArchiTAG, I assist Architect-Professor George Guida to study how AI is reshaping architecture and design intelligence. We explore new tools, emerging ideas, and the evolving role of designers in an AI-driven world. I supported him in research work for preparing lectures such as AI 101 for MDE Harvard 2025 and a Design Intelligence series for M. Arch 2025 students, University of Pennsylvania.
My work involved researching trends, studying real projects, and turning these insights into clear teaching material and hands-on exercises that helped students build their first AI-powered design interfaces.
Exercise 1 | Parametric 3D Tower Builder
Exercise Description: Develop a parametric 3D tower using Vibe Coding and follow it through to AI-enhanced visual outputs.
Objective: To help students learn web development and the use of APIs while exploring AI image generation and the creation of parametric design tools
Interactive controls that let users define the skyscraper’s geometry, proportions, and overall form in real time.
A dedicated panel for capturing, exporting, and organizing visual outputs generated during the design process.
Integrates with the FAL API, allowing users to link image-generation models, Google Nano-Banana, Ideogram, OpenAI, and more, to convert gallery screenshots into high-quality renders
A gallery panel that captures parameter-driven tower design screenshots from multiple angles, allowing users to review them and generate enhanced renders or diagrams using prompts
Generated through AI-driven ideation applied directly onto the building geometry within the Parametric 3D Builder
Tools
Figma Make: To conceptualize and build the initial prototype
Cursor: To take that prototype into code and apply precise, high-impact refinements.
Tech Stack
Three.js
Vanilla Javascript
Fal AI
Exercise 2 | Spatial Comprehension using LLMs
This exercise was developed using the framework presented in the research paper Spacify: A Generative Framework for Spatial Comprehension, Articulation, and Visualization using Large Language Models (LLMs) and eXtended Reality (XR) (2023) by Vaidhyanathan, Vishal; T. R., Radhakrishnan; and Garcia del Castillo Lopez, Jose Luis.
Exercise Description: Develop a web interface that transforms interior floor plans into interactive 3D spaces, allowing users to edit the layout using natural language
Objective: To enable students to learn modern web development while leveraging existing frameworks to build applications powered by LLM APIs
The uploader panel accepts user input in the form of a floor plan image
Scene Info panel maps the layout, including room sizes and furniture counts, and keeps this information updated as the 3D layout evolves with each user query.
The Design Assistant panel is a chat interface powered by the OpenAI API that lets users interact with and edit the 3D layout generated from the floor plan using natural language
Tools
Cursor: To build and refine the prototype with precision
Tech Stack
Three.js
Vanilla Javascript
Flask (Python)
OpenAI API