Clarion

What I built

• Trained an audio event classifier to a 91.5% weighted F1-score on 3,435 validated data points, deployed on an ESP32-S3 with a React Native companion app, running fully offline with no cloud dependency.
• Compressed a 300 MB model to 80.7 KB INT8 — a 99.97% reduction — through knowledge distillation, with quantitative analysis of the accuracy trade-off at each step.
• Awarded a Distinction for the full dissertation — problem definition, literature review, system design, implementation and evaluation.
• Designed the multimodal alert pathway around deaf students' real constraints in Zimbabwean classrooms, where reliable connectivity cannot be assumed.