Hey Ph! i am Dana PhD researcher in AI and architecture, and the person behind Gaia (btw: generative AI architect, and earth goddess :: we need to design better, faster). I have created many products in my life. But Gaia is my favorite. She is a passion, combining work and fun in the right balance.
I built the first version with my co-founder Phil Parvaneh (<3) in '23 at Cornell Tech where I studied computer science with a specialization in urban technology and AI. We had one of the first diffusion models trained on architectural output, tested by architects - won Cornell Startup AwardsGot real customers, then had to pivot due to some operational challenges.
My day-to-day research is about how AI can simulate spatial experience before a building exists. Gaia was supposed to stay on the ice until the models caught up. A dream where time is everything. And then they caught on, without spending $1MM, competing with foundational model providers. in the end.
Almost every day, builders kept asking why we stopped construction. Some had Gaia in their daily workflow eg “using it for a client pitch yesterday, why isn’t lmk exporting” (quite remarkable they were still using V0 Gaia, the pre-cursor/cloud app we wrote by hand in JavaScript (blood, sweat, tears)). When you listen enough times and the tech is finally ready, you rebuild. So :357 later, endless nights of coding, and some good wake-up calls obviously:: I rebuilt everything from the ground up. Design systems for infrastructure, model tweaking, prompt engineering, all after lab hours.
Camera presets are named after Hélène Binet, Julius Shulman, Ezra Stoller, Peter Zumthor — because if you’re going to control perspective, the references have to mean something. This is the kind of decision that a typical AI tool would never make. I tested each preset with 50+ practicing architects, asking: How do you design the best interface between real work and “shiny” AI?
What I am most proud of is the digital twin. Each session is called a process extract_style_dna Creates a model of your design sensitivity. The material, the quality of the light, the composition you keep passing. Over time, Gaia stops needing you to explain herself. Every other tool on the market resets every session to zero. The model compounds, and feeds into, a knowledge graph where each renderer is a node, edges are semantic relationships, and the shape of the graph reflects how your ideas connect. A few architects have told me that they use it like Visual Obsidian for their design thinking.
Also in the toolkit: multi-perspective generation (street, aerial, section cut, interior — one click), reference image stacking with weight input, brush mask editing, PDF export to client decks, and an interior AI with 20+ styles that I know a few hundred architects prefer over the alternatives. is also seeing a wave of homeowners using it to redesign their rooms. Big win, no need to study 6 years?
So Gaia is no longer a startup, no VC, total freedom to build. This is a professional playground for the next few years alongside my PhD. Expect more research-based features. Rather than competing, I’m interested in how and why people actually use these tools. I think the educational angle is a real advantage.
50 free renders to start, or €24/month for unlimited (expenses must be covered). DM me questions or email me at daan@gaia.computer.
Students get 50% towards the end of tuition, and if you’re in one of my current classes, just get in touch and I’ll sort you out.
Don van der Zwaag