2024 · AI pipeline
A vision-model OCR pipeline for restaurant menus
a restaurant technology company · Ai/ML engineer
The problem
A restaurant technology company had a large backlog of menus to digitise, mostly in Turkish and in mixed layouts. Plain text recognition gives flat text with no structure.
What I built
I built a pipeline that sends each menu image to a vision model with a fixed set of fields, then cleans the result, including Turkish characters, into structured data.
The result
A large backlog of menus became searchable data instead of images someone had to retype.
Main tools
- Amazon Bedrock
- Claude
- Python
This is the short version. The full case study is available on request.