(Image credit: Sufyan @blenderdesigner. Unsplash.com)
This five-session workshop provides comprehensive training in contemporary AI deployment methodologies, instructing participants in local Large Language Model execution using frameworks such as Ollama and LM Studio, while facilitating access to open-source models through platforms like AI VERDE. The curriculum encompasses advanced implementation strategies including Retrieval Augmented Generation (RAG) systems for enhanced factual accuracy and hallucination mitigation (Lewis et al., 2020), tool calling architectures for external API integration, and automated text-to-SQL code generation.
Supplemental methodologies include AI-assisted coding techniques, which leverage language models for code completion, debugging, and optimization workflows, enabling accelerated development cycles and improved code quality. Additionally, participants will explore vibe coding approaches—an emergent paradigm emphasizing intuitive, conversational programming interfaces that facilitate rapid prototyping and iterative development through natural language specifications.
The workshop culminates with comprehensive training in agentic systems architecture, where LLMs demonstrate autonomous multi-step reasoning capabilities, strategic planning algorithms, and complex task execution pipelines. These systems represent the current frontier in artificial intelligence applications, enabling sophisticated problem-solving through iterative agent-environment interactions and goal-oriented behavior optimization.
- Lewis, P., Perez, E., Piktus, A., Petroni, F., Karpukhin, V., Goyal, N., ... & Kiela, D. (2020). Retrieval-augmented generation for knowledge-intensive NLP tasks. Advances in Neural Information Processing Systems, 33, 9459-9474. (The seminal paper on RAG).
- Mialon, G., Dessì, R., Lomeli, M., Nalmpantis, C., Pasunuru, R., Raileanu, R., ... & Scialom, T. (2023). Augmented Language Models: a Survey. Transactions on Machine Learning Research. (Provides a broad overview of how LLMs are augmented).
- Mithun, P., Noriega-Atala, E., Merchant, N., & Skidmore, E. (2025). AI-VERDE: A Gateway for Egalitarian Access to Large Language Model-Based Resources For Educational Institutions. arXiv:2502.09651.
- Norman Di Palo and Arunkumar Byravan and Leonard Hasenclever and Markus Wulfmeier and Nicolas Heess and Martin Riedmiller. Towards A Unified Agent with Foundation Models. arXiv preprint arXiv:2307.09668. (Discusses concepts relevant to agentic AI).
Upon completion of this eight-session workshop series, participants will be able to:
- Identify Appropriate Model Applications: Recognize which data types and tasks each architecture best handles (e.g., images, sequences, generative tasks) and understand their ideal application domains.
- Appreciate Implementation Frameworks: Develop familiarity with how these models are built using popular open-source deep learning libraries (e.g., TensorFlow, PyTorch) to support your continued learning and practical applications.
- Deploy and Manage LLMs: Gain hands-on experience setting up and running Large Language Models across different environments, including local instances (e.g., Ollama, LM Studio) and cloud platforms (e.g., AI VERDE).
- Implement Advanced LLM Augmentation: Learn to design and build Retrieval Augmented Generation (RAG) systems that connect LLMs with external knowledge sources to improve response accuracy and relevance.
- Enable LLM-Powered Tool Interaction: Create applications where LLMs seamlessly call external tools and APIs, with particular emphasis on practical applications like text-to-SQL generation.
- Construct Basic AI Agents: Master the fundamental principles of agentic AI systems and build simple agents that can plan and execute action sequences to accomplish specific goals.
Instructors: Enrique Noriega/ Carlos Lizárraga
-
Registration to attend in person or online.
-
When: Tuesday's at 1PM.
-
Where: Albert B. Weaver Science-Engineering Library. Room 212
-
Zoom: (?)
(Program not definitive!)
| Date | Topic | Desciption | Materials | Code | YouTube |
|---|---|---|---|---|---|
| 09/02 | Session 1: Running LLM Locally (Ollama, LM Studio) 💻 | This session introduces the benefits and practicalities of running Large Language Models on local machines. It will cover Ollama, including installation, downloading models (e.g., Llama, Mistral), command-line interaction, and basic API access. It will also explore LM Studio as a user-friendly GUI for discovering, downloading, and interacting with various LLMs, including setting up a local inference server. Brief consideration of hardware requirements will be discussed. | |||
| 09/09 | Session 2: Using AI VERDE (Open-source LLMs) | This session will explore AI VERDE (or a similar institutional platform) as a gateway to accessing and utilizing a curated collection of open-source LLMs. Content will include an overview of the platform's objectives, how to navigate its interface, select different models for specific tasks, and any unique features it offers for research or educational purposes, such as integrated datasets or collaborative tools. | |||
| 09/16 | Session 3: RAG (Retrieval Augmented Generation) 📚 | This session dives into Retrieval Augmented Generation (RAG) to enhance LLM responses with external, up-to-date information. It will cover the core components: document loading and chunking, creating embeddings (e.g., using Sentence Transformers), setting up a vector store (e.g., FAISS, ChromaDB), and the retrieval-then-generation pipeline. The goal is to show how RAG mitigates hallucinations and grounds LLMs in specific knowledge domains. | |||
| 09/23 | Session 4: Tool calling & SQL code generation 🛠️ | This session covers LLM interaction with external tools and APIs. We'll explore function calling in GPT models, teaching participants to define tools, use them effectively, and process outputs. Examples include calculator operations and web search integration. We'll also cover Text-to-SQL generation, showing how LLMs can convert natural language to SQL queries. We'll discuss schema incorporation via RAG and handling different SQL dialects, with hands-on database practice. | |||
| 09/30 | Session 5: Agentic systems 🤖 | This session introduces the concept of AI agents—systems where LLMs are a core component that can reason, plan, and execute sequences of actions to achieve goals. It will cover basic agent architectures (e.g., ReAct: Reason + Act), the idea of an agent loop (observe, think, act), and how agents can utilize tools. An overview of simple agent development using frameworks like LangChain Agents or a conceptual design exercise will be included. |
| Date | Title | Topic Description | Wiki/Slides | YouTube | Instructor |
|---|---|---|---|---|---|
| 01/30/2025 | Scaling up Ollama: Local, CyVerse, HPC | In this hands-on workshop, participants will learn to deploy and scale large language models using Ollama across various computational environments—from laptops to supercomputing clusters—to master practical AI capabilities. | video | Enrique Noriega | |
| 02/06/2025 | Using AI Verde | This practical introduction shows how to effectively use U of A Generative AI Verde for academic research, writing, and problem-solving. Participants will learn to harness AI Verde's capabilities while gaining a clear understanding of its limitations and ethical implications. | video | Nick Eddy | |
| 02/13/2025 | Best practices of Prompt Engineering using AI Verde | A hands-on session that teaches practical prompt engineering techniques to optimize U of A Generative AI Verde's performance for academic and professional applications. | Slides | video | Mithun Paul |
| 02/20/2025 | Quick RAG application using AI Verde / HPC | A hands-on session demonstrating how to build a basic Retrieval-Augmented Generation (RAG) system with the U of A Generative AI Verde API. Participants will learn to enhance AI responses by integrating custom knowledge bases. | Slides | video | Mithun Paul |
| 02/27/2025 | Multimodal Q&A+OCR in AI Verde | A hands-on technical session exploring U of A Generative AI's multimodal capabilities that combines vision and text processing for enhanced document analysis and automated question-answering with OCR technology. | video | Nick Eddy | |
| 03/06/2025 | SQL specialized query code generation | A hands-on session teaching participants how to use Large Language Models to craft, optimize, and validate complex SQL queries, emphasizing real-world database operations and industry best practices. | Slides, Code | video | Enrique Noriega |
| 03/13/2025 | NO Session | Spring Break | |||
| 03/20/2025 | Function calling with LLMs | There are two ways to implement function calling with open-source large language models (LLMs). When an LLM doesn't natively support function calling, you can combine prompt engineering, fine-tuning, and constrained decoding. | video | Enrique Noriega | |
| 03/27/2025 | Code generation assistants | Large Language Models (LLMs) now serve as powerful code generation assistants, streamlining and enhancing software development. They generate code snippets, propose solutions, and translate code between programming languages. | video | Nick Eddy |
| Date | Title | Topic Description | YouTube | Instructor |
|---|---|---|---|---|
| 09/05/2024 | Hugging Face Models (NLP) | Hugging Face offers a vast array of pre-trained models for Natural Language Processing (NLP) tasks. These models cover a wide spectrum of applications, from text generation and translation to sentiment analysis and question answering. | video | Enrique Noriega |
| 09/12/2024 | Hugging Face Models (Computer Vision) | Hugging Face has significantly expanded its offerings beyond NLP to encompass a robust collection of computer vision models. You can find pre-trained models for a wide range of tasks, from basic image classification to complex image generation. | video | Enrique Noriega |
| 09/19/2024 | Hugging Face Models (Multimodal) | Hugging Face offers a diverse range of multimodal models, capable of processing and understanding multiple data modalities such as text, images, and audio. These models are at the forefront of AI research and development, enabling innovative applications. | video | Enrique Noriega |
| 09/26/2024 | Running LLM locally: Ollama | Ollama is an open-source platform designed to make running large language models (LLMs) on your local machine accessible and efficient. It acts as a bridge between the complex world of LLMs and users who want to experiment and interact with these models without relying on cloud-based services. | video | Carlos Lizárraga |
| 10/03/2024 | Introduction to LangChain | Langchain is an open-source Python library that provides a framework for developing applications powered by large language models (LLMs). It simplifies the process of building complex LLM-based applications by offering tools and abstractions to connect LLMs with other data sources and systems. | video | Enrique Noriega |
| 10/10/2024 | Getting Started with Phi-3 | Phi-3 is a series of small language models (SLMs) developed by Microsoft. Unlike larger language models (LLMs) that require substantial computational resources, Phi-3 models offer impressive performance while being significantly smaller and more efficient. | video | Enrique Noriega |
| 10/17/2024 | Getting started with Gemini | Gemini is a large language model (LLM) developed by Google AI. It's designed to be exceptionally versatile, capable of handling a wide range of tasks and modalities, including text, code, audio, and images. This makes it a significant advancement in the field of artificial intelligence. | video | Enrique Noriega |
| 10/24/2024 | Introduction to Gradio | Gradio is an open-source Python library that allows you to quickly create user interfaces for your machine learning models, APIs, or any Python function. It simplifies the process of building interactive demos and web applications without requiring extensive knowledge of JavaScript, CSS, or web development. | video | Enrique Noriega |
| 10/31/2024 | Introduction to RAG | Retrieval-Augmented Generation. It's a technique that enhances the capabilities of Large Language Models (LLMs) by combining them with external knowledge sources. | video | Enrique Noriega |
| 11/15/2024 | Dense Passage Retrieval | video | Mithun Paul |
Created: 06/10/2024 (C. Lizárraga)
Updated: 06/18/2025 (C. Lizárraga)
