Fluently AI: Interactive Language Learning with On-Device OCR & Voice AI
Developed an interactive language learning app combining Google ML Kit camera OCR for real-world vocabulary capture and Whisper voice analysis for pronunciation grading.
The Engineering Challenge
Real-time camera text recognition had to operate offline on low-end smartphones without lagging the live camera preview feed.
Audio recording and pronunciation analysis required low-latency audio capture and synchronization with speech models.
What I Built & Delivered
Integrated Google ML Kit Text Recognition to scan vocabulary from physical books and signs in under 100 milliseconds.
Engineered an interactive audio recording and waveform visualization widget with instant Whisper API scoring.
Built a spaced-repetition memory engine (SM-2 algorithm) that surfaces review flashcards at scientifically optimal learning intervals.
Created engaging gamified quiz interfaces with micro-haptics and animated progress feedback.
Architectural Highlights
On-Device ML Kit Pipeline
Direct camera image stream piping to native OCR C++ libraries without bridge serialization bottlenecks.
Spaced-Repetition Algorithm (SM-2)
Local algorithm tracking recall accuracy scores and scheduling future flashcard repetitions.
Engineering & Business Results
Sub-100ms camera text recognition running completely on-device.
28% increase in 30-day user retention driven by the automated spaced-repetition card algorithm.
Seamless in-app subscription flow converting freemium learners into paid members.
Explore Mobile AI & LLM Integration
Have an app idea similar to Fluently AI?
Let's turn your concept or designs into a live, production-ready Flutter application in 20–30 days.