SproutComm
Built at Meta Global AI Developer Hackathon · Oct 26, 2026 · Remote

SproutComm is a two-way family-engagement system for Head Start programs, built for the families programs reach least. The problem: federal rules require programs to engage every family in their home language — and document it. Our Detroit pilot partner serves 149 children; only 17 come from English-speaking homes. For the other 132 families, engagement today means phone tag, paper flyers that may never be read, and documentation written from memory at the end of the week. How it works: an advocate writes once, in English. The parent receives the message in Arabic, Bangla, or English — as text and as voice, because a message a parent can't read isn't engagement. That night, the parent does a short guided home activity with their child and answers with a voice note, then taps "we did it" plus who joined and how long it took. SproutComm then writes the engagement record itself, composing the send, the voice reply, and the completion into a documentation note in the family file that meets Head Start Program Performance Standard 1302.50. The conversation is the documentation. Each family file also carries a documentation log, a referral tracker, and a printable family report. The parent side is low-literacy-first and audio-first, works offline, and switches fully into the parent's language, including right-to-left Arabic. Built during the hackathon (October 27 – November 4, 2026) from two validated concepts, ParentComm and ReadSprout, SproutComm runs entirely on Meta's models: Muse Spark translates the messages and the full parent interface, Muse Voice Transcribe turns parents' voice replies into text, and Muse Wonder composes the compliance note. The demo runs on synthetic families only — no real family or child data anywhere. Live demo: https://kshepherd0128--sproutcomm-spike-serve.modal.run Site: https://sproutcomm.com