Deep Architecture

Deep Architecture is an experimental project that asks a simple question: Can mass housing still feel tailor-made?
The method harnesses deep learning and natural language processing to read free-form texts—personal diaries, crowdsourced surveys, academic papers—and distils the latent planning principles that link architecture to individual aspirations.

At first, I explored the jump from raw text to built form at multiple scales. Personal notes became a single-family dwelling; public surveys informed a shared building; academic writing generated a neighborhood and, ultimately, an urban proposal. Each step followed the same chain: text analysis → AI image generation → dimensional modelling, while deliberately freezing the viewpoint in a two-dimensional “architect’s eye” to keep the process legible.
I sought a real-world scenario that demands design for the many, yet shares a common cultural denominator. The current state of Israel’s Beta Israel absorption centers became that case study.
The common program of Observation centers in Israel was the one. It allows 6 months of living space for the immigrant to integrate into the country while providing him with a dwelling based on the number of family members. In addition, the program includes public spaces for immigrants to interact and learn the local language, as well as a food court.
The goal: let every new immigrant craft a personal living environment by writing a short paragraph about the place where they last felt “at home.”
Achieving a fully formed design output required the development of new typologies. These were extracted from more than a thousand academic documents on Beta Israel culture, Gondar’s urban fabric, and Ethiopian Jewish heritage.
Just as every newcomer brings a story, the host place brings its own. I interpreted that host through the public spaces building of the absorption center classrooms, dining hall, multipurpose spaces rendered in the vernacular spirit of Be’er Sheva’s (the center’s city) architecture, grounded in scholarly sources and site photography.
The result is a dialogue between algorithm and memory, between collective infrastructure and intensely personal space a proposal for architecture that is both deeply individual and broadly scalable.