A hold-to-talk AI persona of me — Whisper fine-tuned with LoRA and exported to CTranslate2 for real-time int8 transcription, driving a LangGraph agent grounded with RAG over personal data in ChromaDB, with prompt-injection guardrails gated by a CI eval harness.
PyTorch / LangGraph / Whisper / RAG / ChromaDB / HuggingFace
↳ Demo Recording





