Music Intelligence Pipeline

Emotion-aware music analysis system that predicts arousal and valence from audio, then tags music semantically to prevent hallucination and provide an interpretable interface over raw predictions.

• Dataset: DEAM (Database for Emotional Analysis in Music)
• Features: librosa, OpenSMILE
• Model: XGBoost for arousal/valence prediction
• Interface: Shiny for Python (interactive querying)

Raw predictions (R² ~0.43) are directional, not precise. The semantic layer interprets them into broad emotional tags: uplifting, introspective, and similar.

Neural Bard: Spotify LLM Playlist Generator

Full-stack application that generates Spotify playlists from natural language via LLM (Grok), then creates them through OAuth.

• Frontend: HTML5, Bootstrap 5, vanilla JavaScript
• Backend: Netlify serverless functions
• LLM: x.ai Grok-3-fast
• Features: duplicate detection, smart playlist management, OAuth flow

Aimpoint Digital LP Challenge

Mixed-integer linear programming applied to a vehicle routing dataset using scipy.optimize.linprog with the HiGHS solver: constraint design, solver selection, and practical optimization workflow.