Vibe fashion

Vibe Fashion is an AI Fashion Assistant that is a multimodal recommendation system. It delivers personalized fashion suggestions for specific occasions through image uploads and text prompts. Our business model expands into the B2B sector through a REST API, allowing our service to be easily integrated into various fashion marketplaces, helping companies grow their revenue. France is the perfect place to start!

The project meets all core requirements by deploying LLMs (gemma3:12b and Gemini Flash 2.5) to Google Cloud Run with NVIDIA L4 GPU acceleration, and implementing a multi-agent architecture that orchestrates vision and language models for context-aware recommendations. Tech stack is demonstrated through Python/FastAPI REST API implementation with async processing, comprehensive error handling, and full API documentation enabling continuous integration.
The system serves a dual-deployment strategy combining open-source (Gemma3) and commercial (Gemini Flash 2.5) models for optimal cost-performance balance.
The user experience features an intuitive React.js web interface with
1) real-time feedback
2) natural language outfit descriptions
3) visual product suggestions that guide customers from upload to purchase.

Our tech stack includes Gemma3-4b, Gemini Flash 2.5, Cloud Run with NVIDIA L4 GPUs, Docker for containerization, Python, FastAPI, and React.js for the frontend. Scaling and reproducibility are achieved through Cloud Run’s auto-scaling capabilities, response caching mechanisms, horizontal load balancing, fully containerized Docker architecture, and infrastructure-as-code deployment scripts.
The presentation demonstrates end-to-end functionality from image analysis to product recommendations, showcasing real-world applicability for both B2C and B2B fashion e-commerce scenarios.

Our tech stack includes Gemma3-4b, Gemini Flash 2.5, Cloud Run with NVIDIA L4 GPUs, Docker for containerization, Python, FastAPI, and React.js for the frontend. Scaling and reproducibility are achieved through Cloud Run’s auto-scaling capabilities, response caching mechanisms, horizontal load balancing, fully containerized Docker architecture, and infrastructure-as-code deployment scripts. The presentation demonstrates end-to-end functionality from image analysis to product recommendations, showcasing real-world applicability for both B2C and B2B fashion e-commerce scenarios.