AI-435: GenAI Application Development with LLMs (extended version)
(OpenAI GPT, Google Gemini, Anthropic Claude, Meta Llama, Mistral, Deepseek)
Course description: The fast pace of development in LLMs and related technologies made it possible to use them reliably even in enterprise grade applications. There are already a few areas where a new generation of LLM-based applications totally redefined applications’ capabilities and users’ expectations while AI technologies are going to radically change all kinds of other software areas as well.
Consequently, software developers of all tribes need to understand these technologies and need to have practical skills to use them in their daily work.
Training objectives: At the end of the training participants:
• get to know the typical types of LLM-based applications,
• know the basic building blocks, operation and multi-step training of modern LLMs,
• can write simple programs using closed- and open-source LLMs via their own API or through Langchain, the most popular open-source LLM development framework,
• understand the main concepts and most important practical rules of prompt engineering,
• understand the concepts behind RAG systems and can use their basic and more advanced forms in their LLM-based applications,
• understand the concepts of LLM-based Agentic Systems and can write simple autonomous agents,
• understand the reasons for using multiagent systems and can use the Langgraph open-source agentic framework to write and debug simple multiagent systems,
• understand the importance of tracing and evaluation of LLM-based applications throughout their entire lifecycle and can use tracing and evaluation tools such as Langsmith,
• understand the reasons and use cases when fine-tuning open-source LLMs helps achieving better results as well as understanding the technologies behind fine-tuning such as PEFT and quantization.
Main topics:
• Introduction to LLM based applications
• Main parts, working and training of LLMs
• Using closed- and open-source LLMs via APIs
• Creating LLM chains with LangChain
• Fast Web Interface Prototyping for LLMs
• Prompt engineering
• Retriever Augmented Generation (RAG)
• LLM-based Agentic Systems
• Multiagent systems and frameworks
• Tracing and Evaluating LLM-based apps
Besides learning about LLM concepts, students will also do extensive lab exercises using the Python APIs of popular closed-source OpenAI GPT, Google Gemini, Anthropic Claude as well as open-source Meta’s Llama and Mistral models LLMs to see how these concepts work in practice. During the exercises they use LangChain products such as LangChain, LangGraph and LangSmith, to implement LLM concepts in real world LLM applications.
Course Length: 40 training hours
Structure: 50% theory, 50% hands on lab exercises
Target audience: Software developers and other IT and technical professionals as well as managers with technical backgrounds who want to understand the basic concepts and technologies behind Large Language Models (LLMs) and want to obtain practical skills in LLM application development with the Python APIs of popular closed- and open-source LLMs and frameworks.
Prerequisites: Basic understanding of AI concepts, basic Python programming skills, user experience with ChatGPT or similar chatbots.