Vibeset: AI Music Discovery

The Vibe

Vibeset is an AI-powered music discovery platform that goes beyond genres. It understands energy—matching tracks to your birth chart, current mood, or the vibe you’re chasing.

What I Built

The Intelligence Layer

• LLM-Powered Archetype Generation: Uses Mistral and GPT models to translate astrological patterns into musical characteristics
• Real-Time Discovery: Perplexity Sonar API for live web search of new releases matching generated archetypes
• Hybrid Retrieval: Combines semantic search with audio fingerprinting for intelligent recommendations

The Stack

• Backend: Python, AWS Lambda, DynamoDB
• AI: Mistral Small 2503, GPT-4.1, Perplexity Sonar Pro
• Frontend: Streamlit (prototype), Next.js (production)
• Infrastructure: Serverless architecture with 99.9% uptime

The Challenge

Traditional music recommendation looks at what you played. I wanted to predict what you need—based on cosmic patterns, time of day, and emotional state.

Music discovery should feel like magic, not an algorithm.

The Vision

Results

• 10K+ archetype generations in first month
• Sub-2s response time for personalized recommendations
• Zero hallucinated tracks through validation pipeline

This is my flagship project—the system that taught me how to build AI that understands vibes.