{"type":"video","version":"1.0","html":"<iframe src=\"https://www.loom.com/embed/eb44015201554d11ac841cd1700653ce\" frameborder=\"0\" width=\"1220\" height=\"915\" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>","height":915,"width":1220,"provider_name":"Loom","provider_url":"https://www.loom.com","thumbnail_height":915,"thumbnail_width":1220,"thumbnail_url":"https://cdn.loom.com/sessions/thumbnails/eb44015201554d11ac841cd1700653ce-1e812004e9bff2a0.gif","duration":102.605,"title":"y30 voice AI: an unscripted reminiscence call","description":"An unscripted conversation with y30, the voice AI I'm building for older adults and their caregivers.\n\nThe thing to watch is the turn-taking: it waits through a long mid-thought pause without cutting in, and yields the instant I interrupt it.\n\nWhen the conversation turns to my late mother, it softens its tone on its own.\n\nUnder the hood it's a streaming speech-to-text to LLM to text-to-speech pipeline with deterministic flow control and a safety floor that runs independent of any model, every conversation becomes structured signal a caregiver can act on.\n\nPart of my portfolio: whoischrislam.github.io"}