Hands-On LLM Engineering
Published by Pearson
Gain LLM Expertise: Level up your skills to build and deploy AI solutions with RAG, QLoRA and Agents.
- This practical, hands-on workshop equips you with the skills needed to solve real business problems with LLMs.
- Work directly with Closed and Open-Source models and use techniques such as RAG, fine-tuning, and agentization.
- Learn the tools of the trade, including HuggingFace, LangChain, Weights & Biases and Gradio, enabling you to confidently design and deploy powerful AI solutions that drive innovation and deliver immediate impact to your organization.
The surge of progress in Gen AI this year has been unprecedented. There has never been a more exciting time to be a part of this transformative field. As organizations try to solve business problems with Gen AI, a new role has emerged that combines Software Development with Data Science to deliver commercial impact: the LLM Engineer.
This intensive, hands-on workshop equips you with the models, tools, and techniques to harness the power of LLMs. We will start with a tour of frontier models with some practical examples. We’ll discuss how frameworks like HuggingFace make it remarkably quick and easy to deliver significant impact with LLMs.
We’ll then dive into the models themselves, comparing their performance with different tasks, from writing code to graduate-level reasoning. We’ll examine how to select the right LLM to apply to the problem at hand. We’ll also cover some of the latest techniques for training and inference.
In the final segment, we’ll roll up sleeves, pick some thorny business problems, and code up solutions together, using HuggingFace, LangChain, Gradio, Weights & Biases and more. We’ll use the techniques we covered in the previous segments, including RAG, QLoRA, and agentization to create groundbreaking solutions in a matter of minutes.
We’ll end with a look to the future. With new models from OpenAI and Anthropic on the horizon, what are the new innovations and commercial applications we can envision for 2025 and beyond, and how can we best prepare ourselves and our business to be at the forefront of this progress?
What you’ll learn and how you can apply it
- The frontier Closed and Open-Source LLMs, their strengths and weaknesses, and how to select right model for the task at hand.
- State-of-the-art techniques to maximize the impact of LLMs in your solutions, including RAG, QLoRA fine-tuning, and agentization.
- The tools of the trade: platforms like HuggingFace, LangChain, Gradio and Weights & Biases that allow you to design and deploy powerful AI solutions.
This live event is for you because...
- You are a Data Scientist, a Data Engineer, or a Software Developer excited about the potential for LLMs to have massive commercial impact
- You can level up your LLM proficiency and gain the knowledge you need to deliver business value
- You want to leave class armed with tools and techniques that you can put into practice from day one
Prerequisites
- Prior experience with LLMs is helpful but not essential, as this class is practical
- Intermediate Python knowledge, as a software engineer
- Awareness of Jupyter Notebooks
Course Set-up
- Access to Github repo at https://github.com/ed-donner
- Not required, but recommended: If you have time before the class, try following the README instructions in the repo to set up a local environment. There are troubleshooting tips if needed.
- Not required, but recommended – and full instructions will be included in the README of the github repo well in advance of the class:
- A Google colab account to follow along with the colab examples
- An OpenAI API key to run the examples yourself
- A Hugging Face account
Recommended Preparation
- Attend: Statistics Level II: Regression and Bayesian (ML Foundations Series) by Jon Krohn:
- Attend: Intro to Probability Theory (ML Foundations Series) by Jon Krohn
- Attend: Probability, Level II: Distributions and Information Theory by Jon Krohn
Recommended Follow-up
- Watch: Machine Learning Foundations by Jon Krohn
- Attend: Hugging Face in 4 Hours by Sinan Ozdemir
- Attend: Retrieval-Augmented Generation (RAG) and Agents Using LLMs by Sinan Ozdemir
Schedule
The time frames are only estimates and may vary according to how the class is progressing.
Introduction (20 mins)
- Welcome, intros
- Agenda & Goals – what you will leave with
- Setting you up for success – GitHub & resources
- Testing out Live Event features!
Segment 1: Hands on with Frontier Models (40 mins)
- From Start to State-of-the-Art
- Writing code to use LLMs in 5 ways
- Tools of the trade: HuggingFace, LangChain, Gradio, and more
- Going multi-modal
- Break (10 minutes)
Q&A (10 minutes)
Segment 2: State of the Art Models and Methods (60 mins)
- The models
- The Frontier Closed-Source Models
- The Leading Open-Source Models
- Factors for comparing LLMs
- Benchmarks
- Leaderboards
- Arenas
- Techniques
- RAG
- QLoRA
- Agentization
- Break (10 minutes)
Q&A (10 minutes)
Segment 3: TIME TO CODE! Selecting and applying LLMs to solve real problems (60 mins)
- Strategy
- Applying our strategy to 3 actual business problems, hands-on in Jupyter notebooks with curated datasets
- Parsing, generation and summarization
- Generating code
- A question answering system
- Making commercial impact now and in the future
- Preparing for what’s to come in 2025 and beyond
Q&A (10 minutes)
Finale (10 mins)
- Recap
- Resources to put this into action
- A special extra for those who stayed till the end!!
- Next steps and survey
Your Instructor
Ed Donner
Ed Donner is a technology leader and repeat founder of AI startups. He’s the co-founder and CTO of Nebula.io, the platform to source, understand, engage and manage talent, using Generative AI and other forms of machine learning. Nebula matches people and roles with greater accuracy and speed than previously imaginable — no keywords required. Nebula’s long-term goal is to help people discover their potential and pursue their reason for being. Previously, Ed was the founder and CEO of AI startup untapt, an Accenture Fintech Innovation Lab company, acquired in 2020. Before that, Ed was a Managing Director at JPMorgan Chase, leading a team of 300 software engineers in Risk Technology across 3 continents, after a 15-year technology career on Wall Street. Ed holds a patent for a Deep Learning matching engine issued in 2023, and an MA in Physics from Oxford.