Development of fairness-aware training and fine-tuning methods for tabular foundation models

In today’s world, organizations such as banks, schools, and hospitals rely on artificial intelligence (AI) to help make important decisions, from assessing credit risk to detecting fraud. A new type of AI, called Tabular Foundation Models (TFMs), shows great promise for these kinds of tasks. Unlike traditional systems that require heavy retraining, TFMs can quickly adapt to new situations using only a small amount of data, making them faster and more flexible. However, before these models can be trusted in high-stakes settings, it is essential to ensure they make fair and unbiased predictions. Biased AI systems can unintentionally discriminate against certain groups of people, leading to unfair or harmful outcomes.

This project will focus on building fairness directly into TFMs so they remain accurate while also treating all groups fairly. For Layer 6 AI at TD Bank, this research will provide practical methods and tools to build more responsible AI systems. More broadly, the results will contribute to advancing fair and trustworthy AI in Canada.

Faculty Supervisor:

Ulrich Aïvodji;Samira Ebrahimi Kahou

Student:

Partner:

Layer 6 AI

Discipline:

Computer science

Sector:

Finance and Insurance; Professional, scientific and technical services

University:

École de technologie supérieure

Program: