Wireless Analytics for Retail

Refactored an acquired startup ‘interior analytics’ code into a standard production micro-services application deployment (CI/CD, dev/state/production, testing, monitoring)

initially bluetooth, then wifi

Succeeded – got it over 70x faster.

decomposed the ‘application monolith’ w/ micro-service architecture, added production monitoring of ingest and (Bayesian) model drift and credible intervals.

This was all to gauge shopper presence (and location) via wifi or bluetooth.

initially in AWS, ported to GCP

also did a series of multi-modal experiments correlating wireless signals w/ computer vision, and a calibration method was devised that came close to a patent, but was converted to trade secret