Whisper Turbo
Whisper Turbo was and still is the fastest web implementation of the Whisper ASR model from OpenAI. It's powered by a custom Rust inference engine running WebGPU kernels.
A multimodal approach for diagnosis of cognitive impairment.
Prior to my work on on-device ML, I spent some time doing cognitive neuroscience with a focus on Dementia. We explored integrating information from both fMRI and dMRI to improve the accuracy of cognitive impairment diagnosis. We presented our work at the 2021 Organization for Human Brain Mapping (OHBM) conference.