Guru Learner
AI English-speaking practice platform for francophone learners, with 220+ users across 14+ countries.
Visit site- Role
- Founder / Full-stack Engineer
- Status
- Active, in production
- Impact
- 220+ users in 14+ countries
- Stack
- NestJSTypeScriptReactPostgreSQLTypeORMGoogle GeminiOpenAI RealtimeGoogle Speech-to-TextAWS S3Paystack
Problem
Francophone learners often master grammar but lack real speaking practice. Existing apps focus on text and translation, not on the muscle memory of speaking under time pressure with feedback.
Solution
A SaaS combining speaking challenges, real-time AI conversations, transcription-based feedback, gamification, and progression tracking, designed specifically for francophone speakers transitioning to professional-level English.
My role
Founder and sole engineer. Product strategy, architecture, backend, frontend, AI integration, payments, analytics, and ops.
Technical challenges
Low-latency real-time audio with the OpenAI Realtime API; reliable speech-to-text with Google Cloud; combining Gemini for structured feedback with realtime conversation; building a Paystack-based payment flow with referral, top-ups, and gifting; production observability on a tight budget.
What I learned
Shipping an AI product is mostly engineering discipline around imperfect models, not the models themselves. Latency, retries, idempotency, and clear failure states matter more than prompt cleverness.