Majesska

Context

Consumer app for children, teens and adults who struggle to speak or form sentences. It complements speech therapy without replacing it, giving more chances to practice between sessions.

Problem

Practicing speech takes daily repetition and a patient listener. Existing tools are often text-based, too fast or childish for this audience.

Solution

A voice tutor — four voices to choose from — explains each instruction out loud, waits for the answer with no timer, then gently rephrases and invites the user to try again. Six short activities: pictures, reading, sentences, syllables, small words and time.

My role

Product design, accessible design, full-stack development, AI voice pipeline, subscription and launch.

Architecture

• TanStack Start and React for a mobile-first PWA
• Convex for profiles, sessions and subscriptions
• ElevenLabs for voice, Whisper for transcription, OpenAI for rephrasing
• Agents API Hub as a server-side voice gateway
• Stripe for the subscription with a free trial

Challenges addressed

• Giving time to answer, with no pressure and no timer
• Rephrasing without ever saying “that’s wrong”
• Never treating an adult like a child
• Protecting data: the AI says it is an AI, voice is not stored, no ads

Verifiable results

• Live at majesska.com, with a 7-day trial and no credit card required
• Six voice activities and four tutors to choose from
• Parent area: weekly progress, activity choice, 4-digit code
• Four display modes: calm, colorful, high contrast and dark

Next steps

Expand the activities based on feedback from families.