Cosai - Customer Oriented Store Artificial Intelligence From the physical layout to initial customer interaction, Cosai allows stores to optimize their entire experience.
Inspiration
We took inspiration from Amazon Go. Amazon Go can identify how much time users spend at a specific product, and can easily create inferences for store layout optimizations. Furthermore, the biggest pain point we have with T-Mobile stores is the repetitive process of describing our phone plan, current situation, and any other customer oriented details to every new representative we see.
What it does
- Tracks user movements inside T-Mobile stores to identify "hot spots" and provide insights into optimizing physical store layout.
- Automatically recognizes T-Mobile customers and pushes the customer's information to a store representative's device.
How we built it
Node.js server talking to Microsoft Cognitive Services API and storing results in Firebase Firestore. Android app authenticated using Firebase Authentication observing data in Firebase Firestore.
Challenges we ran into
Azure free accounts are rate-limited to 20 calls/minute/API, which made it more difficult to implement real-time computer vision.
Accomplishments that we're proud of
Fighting through the night. Making a decently responsive ML app with a team of only 2.
What we learned
You can't (necessarily/effectively) slap AI/ML on everything.
What's next for Cosai
Inferential learning. Hoarding data to train.
Log in or sign up for Devpost to join the conversation.