Leveraging Large Language Models for knowledge base management and decision making

This project will help a mining company turn its scattered data—reports, manuals, work orders, and sensor readings—into a smart, searchable “knowledge library.” Using large language models (LLMs), such as GPT, the intern will design a simple map of key mining concepts (equipment, events, hazards, procedures), automatically pull important facts from documents and charts, and build a system that can answer everyday questions with clear sources (e.g., “Why did Pump P-23 fail last week, and what should we do next?”). The system will also suggest practical actions—like adjusting maintenance schedules or alarm thresholds—while keeping a record of where each recommendation came from. For the partner, the benefits are faster, more confident decisions, fewer delays and unplanned shutdowns, improved safety, and better use of existing data. The work creates a reusable blueprint the company can scale across sites, making operations more reliable, transparent, and cost-effective.

Faculty Supervisor:

Daeho Kim

Student:

Partner:

WSP Canada Inc

Discipline:

Engineering

Sector:

Construction and infrastructure; Information and cultural industries; Professional, scientific and technical services

University:

University of Toronto

Program: