AI/ML and systems work — LLMs, voice, pipelines, and agents.
Emotional AI companion for children — production voice stack
Production STT–LLM–TTS pipeline (LiveKit) for real-time, low-latency voice; adopted by families and demoed at a technology fair. Architected inference and audio streaming loop with production LLM deployment experience.
Fine-tuned 7B model on Ready.gov & FEMA data; offline-first crisis guidance
Browser-based, local-first app for accurate crisis guidance. Integrated STT–LLM–TTS; used by organizations for emergency preparedness planning.
Privacy-first AI platform for educators
Lesson plans and camp curricula via AI-assisted code and image generation with custom safety guardrails; 10+ active educators, zero data breaches.
Tuya IoT Hackathon — 3rd place (innovation & impact)
IoT-connected build in a 50+ team hackathon; recognized for innovation and real-world impact.
RAG, prompts, and coding agents in production workflows
~15% LLM token reduction and better accuracy via systematic prompt design. RAG pipelines for retrieval; ~20% faster delivery on scoped projects. Coding agents for automation and rapid prototyping.