Portfolio
Speech & Language AI
Real-time pronunciation coaching across seven languages — built, deployed, and live right now, not just prototyped.
AI Pronunciation Coach
Whisper • Real-Time Audio • Text-Similarity Scoring • GPT-4o-mini Feedback
See It In Action
Feedback & Scoring
Sound-Specific Coaching
Specific, actionable feedback — not just a score.
Progress Analytics
Score Trend Tracking
Every scored attempt is logged, so improvement is visible, not assumed.
Core Capabilities
Language Support
Multilingual Practice
Practice sentence generation and coaching across German, French, Spanish, Portuguese, Russian, Japanese, and Chinese.
Sound-Specific Feedback
Specific, Actionable Coaching
Not a vague score: Whisper’s transcript is compared with the target phrase, and GPT-4o-mini points to the specific sounds to practise next.
Score Analytics
Feature & Score Logging
Feature logging and score-trend analytics, deployed through Docker on Render.
Technology Stack
A production system that scores what Whisper hears against the target phrase and adds AI-generated coaching feedback — not just a static demo.
Frontend
- Streamlit
- Streamlit-mic-recorder
- Real-Time Audio Capture
Backend
- FastAPI
- Uvicorn
- Docker
AI / ML
- Whisper API
- GPT-4o-mini
- NumPy
- SciPy
- pandas
MLOps
- Feature Logging
- Render Deployment
- Streamlit Cloud