A Duolingo-style mobile app that teaches how AI actually works—not how to prompt it, but how it thinks, why it fails, and how to reason with it. React Native + Expo (TypeScript, strict mode), in beta on TestFlight and the Google Play internal track.

21 courses across Foundations, Building, Thinking, and Creating. Seven offline-gradeable exercise types—multiple choice, sort, sequence, match, triage, comparison, decision tree—graded by pure local functions with content authored as JSON. Domain logic carries 100% test coverage; a backend handles auth, sync, and social.

Multi-agent orchestration workspace where families of specialized agents coordinate autonomously. Adaptive model + config selection across OpenAI, Claude, Gemini, Llama, and more.

Agents write A2A notes, generate prompts on-demand and iterate every time. Redis-backed async processing. Background intelligence pipelines. Advanced anti-summary context enrichment system.

Cross-platform memory architecture connecting Instagram, YouTube, Gmail, notes. Built unified context layer through semantic linking and vector embeddings. Synthesized multimodal data from diverse sources and offered cross-platform insights and recommendations.

Core technology now integrated into HeyContext. Enables agents to maintain coherent understanding across platforms and sessions.

Voice AI mental health app with Apple Watch biometric integration. Combines real-time conversational AI and journaling with heart rate and HRV analysis. Swift + Python + Core ML.

Science-backed patterns, post-quantum encryption, and reflections users can use. Demonstrated multimodal signal fusion: linguistic + physiological analysis reveals patterns neither shows alone.