top of page


AI-powered classification of text-based descriptions for the restaurant industry

Evaristo Villaseco, Davood Dar


In the US alone, restaurants waste 25bn pounds of food every year before it reaches the consumers plate and independent restaurants are a large driver of this. In this project we have partnered with Burnt (, whose mission is to help restaurants automate their back-of-house operational flow: recipe management, inventory forecasting, analysis and optimization of costs. Restaurants on average use 5-10 different suppliers with varying ways of describing their supplies. To tackle this, we will employ a large language model (LLM) and fine-tune it to process textual description of the supplies and categorize them into predefined classes. We will train our model on collected and unlabelled real-world data provided by Burnt. This will not only aid in streamlining inventory management and procurement, but will also impact management on food costs, menu item gross product, and budget forecasts.

Screen Shot 2022-06-03 at 11.31.35 AM.png
github URL
bottom of page