Whisper is building a new kind of hearing aid. By far the biggest problem people with hearing loss have is understanding conversation in crowded, noisy environments like restaurants, parties, and the like. The classical DSP algorithms that traditional hearing aids run have difficulty in these situations because the noise has very similar statistics to the signal. Deep learning can do far better. What we do differently is that our hearing aid comes with a small pocket unit which has 1800x the compute of a behind-the-ear hearing aid. This allows us to run much more sophisticated deep learning models to remove background noise.
We launched our first generation hearing aid more than a year ago and now are hard at work on the second generation product which will be smaller and produce even higher quality audio.
The company as a whole is around 60 employees, mostly engineers, and there are four of us on the machine learning team. You would work on every stage of the ML pipeline, from data collection and preprocessing, to designing new model architectures, to deploying trained models onto hardware.
There are so many ways we can use ML to improve hearing aids: beamforming, feedback mitigation, personalization, the list goes on. We're only at the beginning of what we believe we can do.
If interested, please apply on our careers page: https://whisper.ai/careers/