Archii

An AI design copilot that turns photos, dimensions, and requirements into smart architectural plans, layouts, cost insights, and visual renderings—from rooms to townships.

Comment

Inspiration

Creating a space often involves the services of an architect and expensive tools and materials. However, it is difficult for the average person to picture what their space can actually look like and to maximise their small spaces without breaking the bank. Our goal was to make architectural intelligence available to everyone. With Archii, you can take a picture of the space and a few variables to get a fully realised AI-generated design plan. Archii was conceived as the idea of being able to carry an AI architect in your pocket.

What it does

Archii is an AI-based architectural designing copilot. The user can upload an image of the room, input dimensions, and specify preferences including style, budget, and usage. The system can then present the user with an analysis of the space, optimal layout designs, and interior images, all while generating cost-based designs. Furthermore, the service provides interaction using an AI chat, allowing users to interact and develop designs through natural conversations.

How we built it

We developed Archii utilising a multi-layered architecture of AI:

  • Frontend: Flutter application, providing a smooth, visual, and interactive experience
  • Backend: Serverpod for managing users, projects, file storage, and AI workflow orchestration
  • AI Engine: Microservices handling using Python:
  • Room and object understanding through computer vision models
  • Layout planning and designing with LLMs
  • Interior Render Generation using Diffusion Models
  • Cost Estimation using Rule-Based and AI Logic All modules are connected using structured APIs, enabling the system to flow progressively from the following: image, analysis, layout, render, cost.

Challenges we ran into

  1. Converting unstructured room images into structured spatial representations
  2. Making AI-generated Layouts Practical, Not Merely Pretty
  3. Keeping design suggestions with realistic budget constraints
  4. Orchestrating several AI models in a suitable order
  5. Creating a UI that is intuitive yet masks complex processes of AI

Accomplishments that we're proud of

  1. A Complete Design Pipeline for AI, not just a Single Model
  2. Development of a System that Integrates Computer Vision, LLM Reasoning, and Generative AI
  3. Organising both functional outputs (layout, cost) and visual outputs (renders)
  4. Defining an experience where the user can “talk” to an AI architect
  5. Organising the project on a scalable model - single rooms, and then further on an expanded level in the future

What we learned

  1. The strongest form of AI comes from multiple models working together.
  2. Prompt engineering is equally important as the selection of a model
  3. Real-world factors such as space and budget are important considerations in applied AI
  4. UI/UX design plays a major role in the user-friendliness of advanced AI
  5. Splitting a large AI problem down into smaller services makes it more manageable to implement

What's next for Archii

  1. Support for full apartment and multi-room planning
  2. AR-based room visualisation
  3. Material and furniture marketplace integration
  4. More advanced structural simulations and lighting simulations
  5. Personalised design style learning based on user history

Built With

Share this project:

Updates