Case Studies
Portfolio
Work that delivers results
Real problems solved with real technology. No fluff.
AI Fitness Tracking System
A cross-platform mobile application using computer vision and pose estimation to track workouts, count repetitions, and provide personalised AI coaching.
Challenge
Real-time pose detection with high accuracy on low-end Android devices.
Solution
Custom TensorFlow Lite model optimised for mobile inference with MediaPipe integration for sub-50ms landmark detection.
Results
- 91% rep-count accuracy
- 40% user retention improvement
- Launched in 3 months
Virtual Makeup AR Platform
A browser-based augmented reality platform enabling customers to virtually try on makeup products in real time using their device camera.
Challenge
Accurate facial landmark tracking across diverse skin tones and lighting conditions.
Solution
Updated TensorFlow.js face mesh model with custom colour calibration pipeline and WebGL-accelerated rendering.
Results
- 23% tracking accuracy improvement
- 40% sales uplift post-launch
- 95% positive user feedback
Offline Navigation App
A cross-platform mobile app delivering full offline maps, turn-by-turn routing, and points of interest search without any internet connectivity.
Challenge
Storing and serving detailed maps for 50+ regions offline within device storage constraints.
Solution
Mapbox offline SDK with SQLite tile cache, delta-update system, and compressed vector tiles reducing storage by 60%.
Results
- Works 100% offline
- 60% smaller map bundles
- 4.8 stars App Store rating