DeepSpark was a solution devised and developed by Habitat Learn for an Innovate UK project, wherein Automatic Speech Recognition (ASR) technologies were to be experimented with. Focusing on taking Speech to Text models and running them on portable devices such tablets, phones, and Raspberry Pi's.
I was a core part of the research and development process, from getting ASR models to run natively on mobile phones offline, and the implementation of web based models for speech to text purposes, along with offline webkit based applications.
This involved fine tuning ASR models, and experimenting with them to ensure efficiency allowed for reasonable (below 5s) response times from less powerful devices.
