Noa Almog
I started writing code for an entirely unromantic reason: I was tired of opening sessions with "so, how was the week?" and getting an answer neither of us actually knew.
Professional background
This section is where a real practice lists its qualifications — degree and institution, licence number and issuing body, years in practice, additional certifications, and professional memberships.
It is deliberately left empty on this prototype. Publishing clinical credentials that cannot be verified against a public register of licensed practitioners would be misleading, so nothing has been invented to fill the space.
How I ended up here
For my first few years in clinic I was convinced the problem was motivation. I sent practice sheets home, explained to parents why it mattered, and when nothing had been done I assumed there had been no time.
It took me far too long to realise I had it wrong. The parents were trying — they simply could not tell whether their child had produced the sound correctly. And when you do not know whether you are doing it right, you stop doing it. That is not a motivation problem. It is an absence of feedback.
I started with voice notes. Parents sent recordings, I listened in the evenings. It worked beautifully for five families and collapsed entirely at twenty. At some point it became obvious that the part I was doing there — listening to one syllable and identifying where it broke down — is precisely the kind of work a machine can do instead of me, freeing me for the part only a person can do.
What has not changed
I still set the plan, listen to the child, and decide the next step. The tool makes no clinical decisions — it counts, measures and flags. When a clinician hands her judgement over to a screen, both sides lose.
What has changed
I walk into a session already knowing what happened across the preceding seven days. It sounds like a small thing. In practice it changes every session.
How I work
Four things I say to parents in the first session, so nothing comes as a surprise later.
Few goals, not many
Two or three at a time. A plan that tries to fix everything at once moves nothing, and turns practice into a burden nobody can sustain.
If nothing is moving, I will say so
We stop and review every six weeks. If it has not shifted we change approach, or I refer you to someone else. I do not keep families in therapy out of convenience.
Parents are part of the team
Not spectators. You will be told what to correct and, more importantly, what not to — because over-correction does more damage than saying nothing, particularly with stuttering.
I say what I do not know
Including about the tool I am building. It is not finished, there are places it gets things wrong, and you will know exactly what they are before you start using it.
Not doing this alone
I lead the clinical side. The engineering is built by people who do it better than I do — and that was a condition, not a compromise.
Speech processing
Research input on speech recognition and phonetic transcription, including regular challenges to the model's assumptions.
Privacy and security
Guidance on building the architecture around recordings of children — local processing, real deletion, minimum collection.
Clinical review
Fellow clinicians who blind-review the engine's output and mark every case where it disagrees with a trained ear.
Sound like a fit?
A 20-minute intro call, free and with no obligation. Even if we conclude this is not the right route, you will leave with a clear answer.