Hestia

Helping imagine your dream home, and communicate it to completion.

Comment

Inspiration

Every homeowner has a dream home in mind, but translating that vision into architectural plans is painfully slow, expensive, and inaccessible without professional help. We wanted to close the gap between a client’s imagination and an architect’s blueprint letting people co-create their dream spaces using AI.

The idea came when one of our teammates experienced first hand someone they knew struggle to explain what they wanted to an architect. The point of Hestia is, what if an AI could instantly visualize it and even draft the floor plan?

What it does

By utilizing Hestia, named after the goddess of the hearth, home, and hospitality, our platform embodies her essence. It creates warmth, safety, and connection in every digital interaction. Hestia helps users generate first an initial few images of how they would want the exterior and main layout would look based on their answer to precisely chosen questions. Based off of this input, and as a final design is selected, Hestia outputs floor plans and a rich description with verbiage appropriate for architect interpretation into a final creative product in the form of the end user’s home.

Whether it’s optimizing energy usage, tracking household tasks, or connecting with local services, Hestia turns everyday living into a smoother, smarter, and more human experience.

How we built it

We built Hestia with the following tech stack:

  • Frontend/backend: NextJS
  • Generative AI: Google Gemini via @google/genai library
  • Database: MongoDB
  • Object storage: Cloudflare R2

Challenges we ran into

We, as per usual with GenAI, had some issues when it came to refining generated images and ensuring context was correct and having proper prompts to ensure high quality generation of assets. We also ran into some issues with image generation model rate limits being incredulously low, billing/fee trial issues, and it misbehaving in what it returned format wise (retuning a floor plan early/SVG instead of the expected PNG or vice versa). There were also momentary issues with image generation latency causing timeouts leading to the need for streaming images as they were generated, leading to more moving parts prone to breakage.

Accomplishments that we're proud of

We were surprised we got image generation eventually working as smoothly as it turned out; with all the hiccups involved, we didn’t expect for it to frankly work as well as it turned out working. We got our full basic user flow implemented even within this time constraints of 36 hours, and I’d say we’re pretty proud of that!

What we learned

  • Synchronizing three different AI calls (Gemini → Diffusion → Floor-plan) under hackathon time constraints.
  • Generating floor plans that look realistic yet are algorithmically built from user text.
  • How small design details like button placement or feedback speed can influence usability.
  • How to design systems that translate creativity into structure.

What's next for Hestia

3D / AR visualization — walk through your AI-generated home in real time. Architect marketplace — share finalized plans and get professional quotes instantly. Smart budget estimator — predict cost from materials and size. Expand dataset to include regional architectural styles (Indian traditional, Scandinavian minimalist, etc.).

Built With

Share this project:

Updates