For higher accuracy, we use two sets of kalman filters to clean the measurements from possible outlier
— Using 17k previous measurements and their real-life errors.
— Using a Multigaussian Variate
distribution, trained with a neural network feed with over 80k body measurements.
We generate a unique set of products for each user, managing millions of database entries.
developed an algorithm for finding user match in an O(log(n)) time complexity.