Desmond Choy

Real-time multi-agent AI music-jamming band with four persistent LLM agents coordinating over WebSockets.

• —6.2s p95 latency across all agent responses
• —100% success rate across 24 directive runs
• —243 commits, 132 issues tracked, 281 tests
• —3 architectural versions shipped in 27 days
• —Custom MCP server bridging live coding environment with agent orchestration

AI-native interactive storytelling app for children that weaves educational content into 10-chapter adventures.

• —Multi-model cost routing: narrative via Gemini Flash, formatting via Flash Lite
• —Validate-then-stream: pre-validates AI content before real-time streaming
• —Visual consistency via accumulated character state + two-step image synthesis

AI-powered PDF study assistant that transforms documents into personalized, adaptive learning notes.

• —Two-phase PDF processing pipeline reducing API costs 70–80%
• —LaTeX/Markdown support with KaTeX math rendering
• —Inline AI commands: elaborate, simplify, analogize

Emotion-aware music recommendation fusing facial expression and scene analysis.

• —Dual-pathway face + scene emotion fusion with variance weighting
• —~19% improvement over scene-only baselines
• —Face detection coverage: 86.1% via MediaPipe

Values-alignment journal analyzing whether daily behavior reflects stated priorities across ten Schwartz dimensions.

• —VIF: ordinal MLP heads with MC Dropout uncertainty estimation
• —Synthetic data generation: 204 personas, 1,651 journal entries
• —Automated LLM judge labeling pipeline via Claude Code subagents

Research on whether fine-tuning shifts self-supervised attention toward semantic features experts consider diagnostic.

• —6 ViT architectures evaluated across frozen vs. fine-tuned conditions
• —Delta-IoU significance testing framework
• —Interactive React dashboard for attention map exploration