Deploy and Query a Small Language Model

IT leaders evaluating AI usually get a stack of vendor claims and very little firsthand evidence. This lab replaces the pitch with direct experience: deploy a small language model (SLM) on your own infrastructure, work through a real configuration problem the way an engineer actually would, and then put the model to work against the exact kinds of tasks a typical IT organization handles every day, from triaging support tickets to summarizing meeting notes to answering questions grounded in internal policy documents.

Because the model runs entirely on CPU with no cloud dependency, this lab also demonstrates a concrete, lower-cost path to on-premises AI, one your organization can evaluate on its own terms rather than a vendor's.

Audience: IT leaders, architects, and technology decision-makers evaluating whether small, on-premises language models are viable for their organization. No hands-on AI/ML background or coding experience is assumed.

By the end of this lab, learners will be able to:

1. Deploy a named small language model to a pre-configured environment using a guided interface.
2. Verify a successful deployment by querying the model's API, testing its capabilities across several realistic enterprise tasks, and validating the responses.
3. Diagnose and resolve a simulated configuration error.

Across five phases (~60 minutes total), learners start an Ollama inference server, discover and fix a realistic network-binding fault, then run the model through four capability challenges (classification, summarization, grounded question-answering, and a deliberate capability-limit probe) before reflecting on where this approach fits their own organization's workflows. This is a hands-on technical walkthrough, not a slide-driven overview — every step is a real command against a real, running environment.

No GPU is required or used; the model runs entirely on CPU, which is itself part of what the lab is designed to demonstrate.