Independent research on speech, fluency and learning, and on AI that people can trust when it matters.
Accra, Ghana · [email protected]
Measuring speech the way a speech lab does, but on everyday hardware: lip and jaw movement from a webcam, plus the sound of the voice. We're building fluent/disfluent classifiers and testing how well they hold up across speakers.
Status: data collection and first models under way. No results yet.
How practice, feedback and spaced review change what people can do under pressure, in interviews, exams and lectures. Our tools log what learners actually do, so questions can be tested against real use.
Status: early.
AI assistants are now used in high-stakes conversations. We're interested in whether their help stays grounded in the user's own material: citations that point to a page, and a visible placeholder instead of an invented fact.
Status: defining our first research question.
We build tools from what we learn, and use them to study it:
We're early and say so. If you work on speech, learning science or AI safety and want to compare notes, write to [email protected].
