CLIMB: Co-operators LLM-driven Interface to MLOps for Business

Machine Learning (ML)-based software quickly becomes complex due to being developed by multiple teams using different programming languages, with documentation stored in various locations. This complexity introduces communication and collaboration challenges, specially due to the fact that Co-operators Business Intelligence department manages over 1,150 assets, which includes ML models, APIs, databases, and more. The current tool for centralizing information, BAM PowerApps, suffers from performance issues, which limits its adoption by developers. Consequently, when maintaining and integrating code into ML projects, developers still face challenges in finding information and documents from other teams. The objective of the project is to compare the User Experience (UX) of using the Co-operators’s BAM PowerApps interface against the proposed CLIMB Chatbot for finding relevant information. This chatbot tool applies advanced AI techniques, such as Large Language Models and Retrieval-Augmented Generation, to find information quickly and easily across multiple document. If the CLIMB Chatbot demonstrates superior usability and faster query resolution, the BI department can streamline operations, reduce time spent search for information, enhance cross-team collaboration, and prevent duplicated work. This efficiency gain will allow developers to focus on higher-value tasks, resulting in improved productivity.

Faculty Supervisor:

Benoit Baudry

Student:

Partner:

Co-operators (Financial Services)

Discipline:

Computer science

Sector:

Finance and Insurance

University:

Université de Montréal

Program: