Ecommerce ops · Production

Handmade.com modernizes product image and description handling with Amazon Bedrock and Amazon OpenSearch Service

The problem

Handmade.com had over 60,000 catalog products with many listings containing basic descriptions insufficient for search and SEO performance. Manual processing consumed on average 10 hours per week and required a team of several people, while sellers expected go-live timelines of under an hour.

Workflow diagram · grounded in source

Image and metadata ingestion

trigger

“Product images and metadata are fetched from the Handmade.com product data repository and Elasticsearch index”

Initial description generation

ai_action

“Anthropic's Claude in Amazon Bedrock is used to generate an initial description for each uploaded image”

Embedding storage in OpenSearch

integration

“The embeddings are stored in an OpenSearch Service vector index, enabling semantic search capabilities”

RAG context retrieval and enrichment

ai_action

“Retrieved context from OpenSearch Service is sent along with the new image to Anthropic's Claude in Amazon Bedrock A refined, enriched product description is generated using the RAG pattern”

SEO metadata generation

output

“Anthropic Claude 3.7 Sonnet is used to generate SEO metadata, including terms and enhanced product narratives”

Behavioral feedback loop

feedback_loop

“The system analyzes user engagement metrics including click-through rates, time-on-page, and conversion events. These behavioral signals are used to refine the prompt engineering for Anthropic's Claude”

Reported outcome

Handmade.com successfully modernized its content generation workflow, streamlining seller interactions and enabling consistent content quality across a large and growing catalog, with vector-based search improving product discoverability and reducing friction in seller content workflows.

Reported metrics

Manual processing time10 hours per week

Catalog sizeover 60,000 products

Dataset sizeapproximately 1 million handmade products

Seller go-live timeline targetunder an hour

Reported stack

Amazon BedrockAmazon OpenSearch ServiceAnthropic's Claude 3.7 SonnetAmazon Titan Text Embeddings V2Amazon API GatewayElasticsearch

Source

https://aws.amazon.com/blogs/machine-learning/how-handmade-com-modernizes-product-image-and-description-handling-with-amazon-bedrock-and-amazon-opensearch-service?tag=soumet-20

Read source ↗

Frequently asked questions

What did this team achieve with this AI workflow?

Handmade.com successfully modernized its content generation workflow, streamlining seller interactions and enabling consistent content quality across a large and growing catalog, with vector-based search improving pro…

What tools did this team use?

Amazon Bedrock, Amazon OpenSearch Service, Anthropic's Claude 3.7 Sonnet, Amazon Titan Text Embeddings V2, Amazon API Gateway, Elasticsearch.

What results were reported?

Manual processing time: 10 hours per week; Catalog size: over 60,000 products; Dataset size: approximately 1 million handmade products; Seller go-live timeline target: under an hour (source-reported, not independently verified).

How is this ecommerce ops AI workflow structured?

Image and metadata ingestion → Initial description generation → Embedding storage in OpenSearch → RAG context retrieval and enrichment → SEO metadata generation → Behavioral feedback loop.