Guillaume Martin

Developed a web application that allows users to create a conversational intelligence—a chatbot—and deploy it on different channels such as a site, Messenger, Google Assistant, Slack, etc.

It was a challenge in terms of ergonomics because the user must be able to define conversation nodes and then visualize the conversation in the form of a tree and its ramifications. The user must also be able to help the chatbot understand specific terms or phrases using language understanding tools such as Dialogflow or LUIS. Finally, he must be able to test his chatbot in a fully customizable and exportable interface.

For this project, I was in charge of designing and implementing the interface, updating the conversational engine back end, which was only a prototype at the time, and connecting the APIs of the different deployment channels and NLP tools.

This web application allows a company to simulate marketing investment scenarios on different channels, including TV and digital, and to visualize its direct impact on sales. The predictive models built with marketing mix modeling are loaded on the platform and can be adjusted to test different hypotheses.

This project had a strong data component. On the back end, I worked on things like storing time series and calculating the model outputs, while on the front end, I worked on things like the input interface for hypotheses, data visualizations to analyze the impact on sales, etc.