LastMile AI Project

Our research aims to address the trustworthniess, scalability and interpretability of LLMs in the RAG enviroment by generating not only novel academic insights but also practical technology that helps enterprise AI users better understand, evaluate, and debug. By achieving a deeper, theoretically and empirically grounded understanding of how LLMs behave within RAG systems, we seek to provide actionable insights directly to users of industrial systems. Current technical approaches to evaluating LLM responses are often limited to coarse-grained classifications, which can be inaccurate, difficult to interpret, and not easily actionable. Our goal is to develop technology that provides users with fine-grained, interpretable, and actionable information, allowing for meaningful improvements to production systems that would otherwise be difficult to achieve.

Faculty Supervisor:

Ga Wu

Student:

Partner:

LastMile AI

Discipline:

Computer science

Sector:

Professional, scientific and technical services

University:

Dalhousie University

Program: