GRDF: Predicting Risk in Construction Sites

The following Q&A occurred during the Everyday AI Conference in Paris.

動画を視る

How Do You Use AI in Risk Management?

Gaz Réseau Distribution France (GRDF) predicts whether there will be damage on work and construction sites or not. We do this by taking a lot of historical data from the work declarations and historical damages and then building an AI model able to predict the risk. We then target our visits on these construction sites to prevent damage and warn the construction companies if they are doing something they should not be doing, while giving them advice.

How Do You Use Dataiku?

Dataiku is a great AI toolbox for GRDF. It helps us with code management and collaborative work between data scientists. It’s a great code environment manager and our toolbox that reduces the time to market for all our experiments by a lot and helps us industrialize and push projects to production.

What Are the Next Steps for GRDF?

Our next step is Generative AI. We have identified a few use cases and we plan on selecting a few of them to start. We’re currently working on identifying all the parts that require work in Large Language Models (LLMs) and determining how to leverage them. The idea is to try, from end to end, to have an LLM that works on our use cases and then change some parts of it until it can answer every use case perfectly. 

What Does Everyday AI Mean to You? 

Everyday AI means helping the business, sometimes without even knowing it, with AI.