Ecommerce ops · Production

Walmart builds Ghotok to improve product categorization using ensemble predictive and generative AI across 400 million SKUs

The problem

With over 400 million SKUs on Walmart.com, items were sometimes mistakenly categorized due to the sheer number of categories and product types, causing less relevant items to appear in customer searches.

Workflow diagram · grounded in source
1
Fetch items by category
trigger
“Fetch items based on the category they fall into”
2
Ghotok identifies product types
ai_action
“To make certain that the most relevant product types are identified in step 2 for each category or subcategory, Walmart employs a state-of-the-art AI technique named Ghotok”
3
Predictive AI confidence filtering
ai_action
“As ML inference using Generative AI technology is costly, we do not run Generative AI models for all the millions of candidate pairs. For each Predictive AI model and for each <Category, ProductType> pair, we first learn confidence thres…”
4
Generative AI relevance filtering
ai_action
“We then use Generative AI methodologies and their learned relevance thresholds to filter out more irrelevant pairs”
5
Filter items by product types
validation
“Filter the items fetched on step 1 based on the product types in step 2”
6
Exception handling in production
human_review
“we developed an exception handling tool, powered by both machine learning and human intervention, which facilitated swift and seamless resolution of these issues”
Reported outcome

The ensemble of predictive and generative AI models showed the best performance on a validation set and successfully lowered the False Positive Rate to a satisfactory level, with an exception handling tool resolving remaining edge cases in production.

Reported metrics
candidate pairs after predictive AI filteringfrom millions to thousands
False positive ratelowered the False Positive Rate (FPR) to a satisfactory level
Ensemble model validation performanceshowed the best performance
Reported stack
Ghotok
Source
https://medium.com/walmartglobaltech/using-predictive-and-gen-ai-to-improve-product-categorization-at-walmart-dc9821c6a481
Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

The ensemble of predictive and generative AI models showed the best performance on a validation set and successfully lowered the False Positive Rate to a satisfactory level, with an exception handling tool resolving r…

What tools did this team use?

Ghotok.

What results were reported?

candidate pairs after predictive AI filtering: from millions to thousands; False positive rate: lowered the False Positive Rate (FPR) to a satisfactory level; Ensemble model validation performance: showed the best performance (source-reported, not independently verified).

How is this ecommerce ops AI workflow structured?

Fetch items by category → Ghotok identifies product types → Predictive AI confidence filtering → Generative AI relevance filtering → Filter items by product types → Exception handling in production.

Related ecommerce ops cases