An OpenHome builder project · Alpha · Summer 2026

A voice assistant that understands when others don't.

Noah is 11. He has Childhood Apraxia of Speech. He understands every word you say. He just can't always make his mouth produce the sounds he means. Alexa gives up on him. Google interrupts him. So we're building something that doesn't.

Join the alpha Read why
Built by Noah's dad
Powered by OpenHome
Clinically informed
The story

Voice tech was never built for him.

Childhood Apraxia of Speech is a motor planning disorder. Roughly 1 in 1,000 kids has it. Their brains know the word. Their mouths can't execute the precise sequence of movements to say it cleanly. The same word can come out differently from one attempt to the next.

Every voice assistant on the market — Alexa, Google, Siri — was trained on the speech patterns of neurotypical adults and tuned for fast, command-style interaction. They time out. They mishear. They interrupt. For a kid like Noah, each failed interaction is a small reminder that the world wasn't built with him in mind.

He stopped asking the speaker for things. So I'm building a speaker that won't make him stop again.

Project Noah started as a single OpenHome dev kit on a kitchen table. It's now becoming a voice AI platform designed from the ground up for kids and adults whose speech doesn't fit the neurotypical template — CAS, stuttering, dysarthria, AAC users, and beyond. Built with speech-language pathologists. Tested by the families who live this.

The approach

Built around how Noah actually talks.

01 / Patience

Long pauses are okay.

Most assistants time out in under a second. Project Noah waits. Silence isn't failure — it's processing. He gets to finish the thought.

02 / Approximation

"Wabbit" still means rabbit.

The model is tuned for consonant substitutions, dropped syllables, and inconsistent productions. Context fills in what articulation can't.

03 / No correction

The assistant never corrects.

Pronunciation work is the speech therapist's job. The assistant just listens, understands, and responds — modeling correct speech naturally without making it a lesson.

04 / Custom wake word

A name he can actually say.

"Hey Google" requires a hard /g/ — exactly the motor pattern CAS breaks. Every family picks a wake word built from sounds their kid produces reliably.

05 / Local-first

His voice stays home.

Runs on local hardware via OpenHome. No cloud dependency required. Voice data doesn't leave the house — important for kids, essential for schools.

06 / Clinically informed

Designed with SLPs.

The conversation logic is reviewed by speech-language pathologists who specialize in motor speech disorders. This isn't an AI demo. It's a clinical tool.

How it works

The technical bet behind Project Noah.

Most accessibility tech is built on cloud platforms that weren't designed for sensitive data, for families on shaky bandwidth, or for schools with strict student-privacy requirements. Project Noah is built on OpenHome's local-first voice AI stack, then layered with a model fine-tuned on disfluent speech and a clinical content library co-developed with SLPs.

Hardware
OpenHome DevKit
Raspberry Pi-based local voice AI device. No Amazon, Google, or Apple in the loop. Open SDK, fully customizable.
Speech model
Fine-tuned for CAS
Trained on consented disfluent-speech samples. Optimized for consonant substitution, syllable simplification, and within-speaker variability.
Privacy
Local-first, PHI-aware
Voice processing stays on-device by default. Designed to clear COPPA, FERPA, and the privacy reviews that block Alexa from schools.
Clinical layer
Co-developed with SLPs
Conversation logic, content library, and progress tracking built in collaboration with speech-language pathologists certified in motor speech disorders.
Join the alpha

Five families. This summer.

We're taking five CAS families into the alpha cohort this summer. You'll get a pre-configured device, a custom agent tuned to your kid, and direct access to the team. In exchange, we ask for honest feedback and permission to learn from how your kid uses it. No fee. No commitment past the alpha. Just real families helping us build the real thing.

Priority for families connected to Apraxia Kids or working with a CAS-certified SLP.
Questions: hello@projectnoah.dev

Frequently asked

For parents and clinicians.

Is this a replacement for speech therapy?

No, and it never will be. Project Noah is built to be a patient conversation partner and home tool that complements the work your child does with their SLP. The clinical advisors on this project have been explicit: AI doesn't replace a trained therapist, but the right tool can extend what kids practice between sessions and make daily life less frustrating.

What does it cost?

The alpha is free for the five families we select. After the alpha, we expect to offer a hardware bundle (the OpenHome device, pre-configured) for around the cost of materials, with an optional software subscription that funds the clinical content library and model improvements. School and SLP-practice pricing will be separate.

What data does it collect, and where does it go?

Voice processing runs locally on the OpenHome device by default. We do not stream audio to cloud servers for transcription. Any data used to improve the speech model is opt-in, anonymized, and family-controlled. We're designing for FERPA and COPPA compliance from day one because schools and pediatric users require it.

I'm a speech-language pathologist. How can I be involved?

We're actively building a clinical advisory group of SLPs certified in motor speech disorders. If you work with CAS kids and want to shape the conversation logic, the content library, or the progress-tracking layer — reach out. Credit, compensation, and real influence on the product.

My child has a different speech disorder. Will this work for them?

The first version is tuned for CAS, but the underlying approach — patient timing, approximation tolerance, custom wake words, no shame on misrecognition — helps anyone whose speech doesn't fit the neurotypical adult template. Stuttering, dysarthria, AAC users, ESL learners. Apply for the alpha and tell us your situation.

Who is building this?

Project Noah is being built by Brennan Decker — Noah's dad, and a product leader who's spent the last decade building trust, identity, and personalization systems at companies like Ticketmaster, Bluecore, Fanatics, and Finish Line. The clinical guidance comes from a speech-language pathologist certified in motor speech disorders, with a formal advisory group of CAS-certified SLPs being assembled now. We're not a VC-backed startup yet. We're a small, focused team building the thing that should exist. LinkedIn ↗