Model Serving

How are machine learning price predictions served to customers?

Barker's machine learning models serve customers via a client portal and API integration that is white labeled in each served marketplace. When a customer consigns an asset to a marketplace, the customer is prompted to provide a set of fields describing the asset.

Take a collectable baseball card for example. The consignor would be asked to provide the player, grade, print year, any present errors, and the number (proxy for scarcity). These features, along with the image, current market conditions, and any other descriptors would then be fed into the model.

Barker's machine learning algorithms would then provide a price warranty valuation for the card. Barker's model serving architecture follows the below API structure, as neatly outlined in the MLOps, industry resource:

Barker uses this model serving architecture to provide easy integration into each marketplace while still allowing for daily model updates to capture fluctuating market conditions.