The University of Kentucky has developed a secure institutional platform that democratizes access to customized large language models while maintaining enterprise-grade security controls.

• Leverages multi-LoRA inference to efficiently support multiple custom adapters across diverse users and projects
• Provides end-to-end capabilities from dataset curation to model training and secure inference
• Implements robust security measures including process isolation, end-to-end encryption, and role-based authentication
• Creates a user-friendly environment for exploring and deploying custom LLMs within institutional boundaries

This research addresses the critical challenge of balancing innovation with security requirements in enterprise AI adoption, offering a blueprint for organizations seeking to implement LLMs while maintaining control over sensitive data and models.

Institutional Platform for Secure Self-Service Large Language Model Exploration