Inspiration
A room can feel cramped, inefficient, or uncomfortable even when there is enough space to make meaningful improvements. The challenge is knowing what to change, where to begin, and whether a proposed improvement is practical or affordable.
Professional interior design advice isn't always accessible, especially for students, renters, and households working with limited budgets. I wanted to explore how technology could make everyday space planning more approachable.
That inspired Spacewise: a room-planning assistant designed to help people understand their existing spaces, evaluate improvement ideas, and make more informed decisions before spending money.
What it does
Spacewise helps users explore practical improvements to their rooms through a combination of visual analysis, structured assessment, and conversational guidance.
Users can upload a photograph of their room and select improvement goals, such as better space utilization or more effective use of their budget.
The application is designed to identify relevant room objects, generate improvement recommendations, and present five assessment indicators:
- Design Confidence: How well the proposed changes align with the selected goals.
- Space Utilization: Opportunities to use the available space more effectively.
- Budget Compatibility: How well the proposed improvements fit budget-related priorities.
- Feasibility Confidence: The estimated practicality of implementing the recommendations.
- Overall Compatibility: A combined assessment of the proposed improvements.
Spacewise also includes a conversational prototype through which users can describe their room-planning needs in natural language and receive guidance.
The longer-term vision is a more interactive experience in which users can refine room designs conversationally rather than repeatedly starting their planning process from scratch.
How we built it
I developed Spacewise using a combination of web technologies, Python, and cloud-based image-analysis services.
The application includes:
- Frontend: HTML, CSS, and JavaScript for the room-upload interface, improvement-goal selection, assessment displays, and conversational interaction.
- Backend: Python and FastAPI for handling requests and coordinating analysis.
- Amazon S3: Part of the image-storage workflow.
- Amazon Rekognition: Integrated into the experimental image-analysis pipeline for detecting recognizable objects in room photographs.
- Gemini: Used in the recommendation-generation pipeline.
- Deterministic assessment logic: Used to produce five structured scoring indicators based on available analysis and selected improvement goals.
I also developed a rules-based conversational prototype to explore how users could express their needs naturally, including requests involving limited budgets and small rooms.
During development, I investigated Amazon Bedrock as a potential foundation for more advanced multimodal analysis and conversational refinement.
The architecture separates image processing, recommendation generation, assessment logic, and user interaction so that individual components can be improved independently.
Challenges we ran into
One of the biggest challenges was making cloud-based image analysis reliable.
Amazon Rekognition repeatedly returned errors during testing, including validation failures that prevented successful image inspection. Troubleshooting required investigating request formats, cloud configuration, image handling, and integration between the frontend and backend.
Another challenge involved the browser's photo-upload control. Although the upload area appeared correctly, selecting a photograph did not consistently open the native file picker. This required debugging the interaction between HTML, CSS, and JavaScript.
Building a useful assessment system was another important challenge. Interior design decisions are subjective, and an automatically generated score should not be presented as an objective guarantee. I therefore used deterministic scoring rules and included explanations and limitations rather than treating the scores as definitive measurements.
Finally, developing and testing multiple integrations within a hackathon timeframe required prioritizing a demonstrable user experience while keeping experimental functionality clearly distinguished from verified functionality.
Accomplishments that we're proud of
I'm proud of turning Spacewise from an initial room-planning concept into an interactive application with a structured workflow.
Some of the most meaningful accomplishments include:
- Building a unified interface for room photographs, improvement goals, recommendations, and assessment results.
- Developing five interpretable assessment indicators rather than relying only on unstructured AI-generated advice.
- Creating a conversational prototype that responds to natural-language room-planning requests.
- Designing the application to account for practical constraints such as affordability and space limitations.
- Exploring how cloud-based computer vision and language models can complement transparent, rules-based evaluation.
As a solo developer, I also gained experience coordinating frontend development, backend services, cloud integrations, and debugging under tight time constraints.
What we learned
This project reinforced several important lessons.
First, integrating AI services involves much more than calling an API. Permissions, service availability, input validation, error handling, and testing all determine whether an integration is genuinely usable.
Second, explainability matters. Users should understand why a recommendation was made and what limitations affect its reliability.
Third, conversational interaction can make a planning application more accessible. People naturally describe problems such as "my bedroom feels too small" rather than specifying technical design parameters.
Finally, a reliable, limited feature is often more valuable than an ambitious feature that has not yet been tested successfully.
What's next for Spacewise
My next priorities are to improve reliability and expand the depth of Spacewise's analysis.
I plan to:
- Stabilize the room-photo analysis workflow and improve handling of different image types.
- Explore Amazon Bedrock multimodal models for richer visual understanding and contextual recommendations.
- Develop more advanced conversational refinement, including potential Alexa+ experiences.
- Add room dimensions, furniture measurements, and more detailed budget constraints to improve assessment accuracy.
- Support comparisons between alternative room arrangements.
- Improve transparency by showing the assumptions, limitations, and reasoning behind recommendations.
- Explore accessibility improvements and more personalized room-planning experiences.
My long-term goal is to make Spacewise a practical decision-support tool that helps people improve their living spaces with greater confidence, regardless of their design experience or budget.
Log in or sign up for Devpost to join the conversation.