Mirror - Multimodal Embeddings Platform

🥇 Winner of Tech Europe Paris AI Hackathon 2025

Mirror is the fastest and most cost effective way to build web apps with multimodal embeddings. Backed by OpenAI CLIP, Weaviate and NextJS (Lovable), build the next Pinterest, Instagram or whatever you could imagine with no GPU’s, no complex deployments and no drain on your wallet.

What We Built

We’ve shipped our frontier UI to demo the capabilities of Mirror, we indexed thousands of images in just a few hours costing less than a few cents and we can enable natural language search over these images forever, only paying for what you use, or if within the free plan, absolutely nothing!

Key Features

• Serverless Solution: No complex deployments, no GPU requirements
• Cost Effective: Index thousands of images for just a few cents
• Natural Language Search: Search through images using natural language queries
• Scalable: Built on Weaviate vector database for enterprise-scale applications
• Open Source: Fully open source and community-driven

Tech Stack

• Frontend: Next.js, TypeScript, Tailwind CSS
• AI: OpenAI CLIP for multimodal embeddings
• Database: Weaviate vector database
• Storage: Vercel Blob for efficient image storage
• Deployment: Vercel platform

The Hackathon Experience

Built with Cody Adam and Jacob Kenney during the Tech Europe Paris AI Hackathon. We focused on solving the core problem of making multimodal AI accessible to developers without the complexity of GPU management or expensive infrastructure.

The jury included industry experts from MakiPeople, OpenAI, ACI.dev, Beyond Presence, Lightpanda, and Dust, who recognized our solution’s potential to democratize multimodal AI development.

Impact

• Winner: 1st place at Tech Europe Paris AI Hackathon 2025
• Open Source: Available on GitHub for community contributions
• Scalable: Designed to handle enterprise-scale image indexing and search
• Cost Efficient: Dramatically reduces the cost of multimodal AI applications

Try It Out

Visit mirror-azure.vercel.app to experience the power of multimodal embeddings, or check out the source code to build your own applications.