The Challenge
Choosing a material before purchase is a leap of faith. Renders look perfect, but only for the options you already picked — never for the ones you're still deciding between. Architects, ceramic suppliers, and individual clients all faced the same problem: no fast, realistic way to compare materials in their actual space before committing to a purchase.
The Solution
I built an AI-powered visualization tool using Stable Diffusion-based models. Users upload a photo of their space, select a material, and instantly see how it would look — applied directly onto the existing image, in real lighting conditions. I validated the concept through Teknopark's pre-incubation program, running structured interviews and tests with 5 architects, 5 ceramic suppliers, and 1 individual client.
Watch It In Action
If the embed doesn't load, view the original post on LinkedIn.
Project Details
Validated with 5 architects, 5 ceramic suppliers, and 1 individual client. The concept didn't reach product-market fit — but the process clarified my direction toward data-driven, technical work.
Development Process
User Research
Interviewed architects and ceramic suppliers to understand the real decision-making pain point before purchase.
Model Development
Built an AI pipeline to apply material textures realistically onto uploaded photos of real spaces.
Validation & Pivot
Tested with real users, gathered feedback, and learned the idea needed a different direction — a turning point that shaped my path toward data science.