Specialty Produce AI

Due to the sensitive nature of internal financial records, the dashboard is currently not available to the public.

Specialty Produce is a family owned and operated food service and retail fresh produce supplier located in San Diego, CA. They supply to local restaurants, business, and are also open to the public. This creates a particularly difficult task to predict volumes as orders can range from a single item to multiple palettes. In order to overcome this, machine learning was implemented in order to provide insights into current-day as well as future orders.

Each product is modeled individually. This is a difficult task as they carry thousand of products, many of them perishable.

Numerous models were created and incorporated into the AI:

  1. Order Quantity Fit
  2. Order Quantity Upper Bound
  3. Order Probability
  4. Order Expiration Quantity
  5. Daily Order Quantity Sellout by Expiration Probability
  6. Expiration Order Quantity Sellout by Expiration Probability

Each of those sets of models are built for

  1. Current-Day
  2. Next-Day
  3. Two Days

This data is all plotted on an interactive chart in order to monitor the status of the models

  1. Point: Actual Order Quantity
  2. Cross: Same-Day Order Quantity Fit Prediction
  3. Colored Bar: Error (Color indicates direction of error)
  4. Dark Grey Bar: Same-Day Order Quantity Fit and Upper Bound
  5. Light Grey Bar: Next-Day Order Quantity Fit and Upper Bound

Models are batched daily. Each model is incorporated into final prediction based on the cost of the product (loss) and the price to rush-order that item (cost). Upper bounds are predicted in order to account for the loss associated with that product in general is less than the cost of rush-ordering additional quantity last-minute.

Artificial Intelligence Consultant,

My focus is on full-stack deployments of AI with end-to-end project development

Related