More than 1 million people in India and 10 million over the world are known to suffer from Parkinson's disorder (we have no idea how many more are yet undetected). That along with with neuro-motor dysfunctions, like Bell's Palsy, ALS (Amyotrophic lateral sclerosis), PLS, PBP (Progressive bulbar palsy) are often detected late because of lack of awareness of detection of early onset symptoms, ignorance and the dismal neurologist-patient ratio of our country.
We cannot replace a doctor's diagnosis, or even try to augment it, but we can monitor late middle age and old age people on a daily basis on multiple paramaters like body mobility (which can detect paralysis, mild strokes, palsy), facial expression diversity (which can detect multiple nervous palsies) and ability to draw shapes .
Tracking these parameters on a daily basis, and detecting anomalies in patterns (someone unable to raise their right hand fully since yesterday or drawing more wavy lines than before) can help us trigger alerts for the primary care givers, their immediate family and/or their doctor, who can intervene with prompt diagnosis and appropriate action.
Realtime image processing on device for Android is already quite difficult. But we also had to tune the models to detect particular keypoints that are needed to detect the symptoms we need.
We also spent quite some time figuring out how to make this fun and less intrusive, so that those who use it won't feel it is like a chore. We finally figured out using 3D characters who crack jokes can be good way to do this.