Furniche

Winner @ 42 Bangkok x AWS Hackathon

Impact

Winning solution that revolutionizes furniture design using a multi-modal Generative AI pipeline on AWS.

Overview

Revolutionizing Furniture Design with Generative AI: Inside the Winning Solution of the 42 Bangkok x AWS Hackathon

We are thrilled to share the technical journey behind Furniche, our project that secured the 1st place trophy at the 42 Bangkok x AWS Hackathon: Future x Smart Hub.

In the traditional furniture industry, the journey from a vague creative vision to a tangible product is often slow and fragmented. Our mission was to build a platform that leverages Generative AI and Cloud Infrastructure to bridge this gap, connecting users, SMEs, and freelancers in a seamless digital ecosystem.

Figure: Furniche Hero

The Architecture: A Multi-Modal AI Pipeline

Our platform is designed as a sophisticated pipeline that moves from multi-modal user inputs to high-fidelity designs and automated business workflows.

Figure: Furniche Architecture

1. Intelligent Input & Prompt Orchestration

A unique design starts with the right intent. Furniche allows users to provide input through two main channels:

Visual Context: Users can upload a "Preference Image" to set the initial style or aesthetic direction.
Voice-to-Prompt: Recognizing that inspiration often comes through speech, we integrated Amazon Transcribe. This service converts spoken descriptions into text prompts, which are then passed to our preprocessing engine.

To ensure high-quality outputs, we implemented a Prompt Pre-processing layer powered by a GPT Service. This layer acts as a "prompt engineer," polishing raw user input into descriptive, environment-friendly prompts that are optimized for generative models.

2. The Generative Engine

At the heart of Furniche lies the Image Generative AI. By combining the reference image with the polished text prompt, the model generates a bespoke furniture design. This isn't just a static image; the system also extracts Design Details—technical specifications and material descriptions—derived from the LLM’s processing of the original prompt.

Figure: Furniche Consult

3. Downstream Integration & Smart Search

Once a design is generated, the platform triggers two distinct workflows:

Visual Search via Amazon SageMaker: The generated image is fed into a custom Product Search AI hosted on SageMaker. This allows the system to cross-reference the design against existing inventories or similar catalog items, providing immediate options for purchase or inspiration.
The Auction Marketplace: The "Design Details" are automatically pushed to our Auction Feature. Here, the project transitions from a design tool to a business platform.

4. Closing the Loop: LINE & Amazon Lex

To facilitate real-world production, we built a notification system to alert the maker community:

Amazon Lex processes the auction details and manages the conversational logic.
LINE Chatbot Integration: Once an auction is live, the system automatically notifies registered SMEs, craftsmen, and freelancers via LINE. This enables artisans to bid on projects that align with their specific skills and production capabilities.

Figure: Furniche Ending

Social Impact: Empowering SMEs and Freelancers

Beyond the technical novelty, Furniche is a tool for economic empowerment. By providing an Auction Market, we lower the barrier for SMEs and freelancers to access high-quality custom design leads. Users get the flexibility of choosing a skilled artisan that fits their budget and project needs, while creators gain a streamlined channel to discover new business opportunities.

Conclusion

The "Future x Smart Hub" hackathon challenged us to think about how technology can make our environments smarter and more collaborative. With Furniche, we’ve demonstrated that by orchestrating AWS AI services and LLMs, we can transform a creative spark into a collaborative marketplace, revolutionizing how furniture is designed, auctioned, and crafted for the modern world.

Generative AIAWSSageMakerStreamlitLINE API

Gallery Overview