Small Language Models are gaining traction as a practical alternative to large, cloud-hosted models. They can run on local hardware, provide greater control over sensitive data, and be adapted efficiently to specialised tasks. This presentation introduces Vision Language Models and shows how they can be deployed on local systems without relying on external APIs. It then focuses on a hands-on case study involving the LoRA fine-tuning of a Vision Language Model for information extraction.

What participants will do

Participants will explore the opportunities and trade-offs associated with running smaller models locally. They will learn how Vision Language Models can be deployed on their own hardware and follow the steps of a case study showing how LoRA fine-tuning can adapt a model to a specific information-extraction task.

What to expect

Participants can expect a practical introduction to local and specialised AI models, with a focus on privacy, efficiency, and task-specific adaptation. The session will connect current model concepts with a concrete application, helping participants assess when a Small Language Model or a locally deployed Vision Language Model may be a suitable choice.