Course Outline
Day 1 – Generative AI & LLM Fundamentals
Overview of generative AI and real-world LLM applications
Exploring transformer architectures (GPT, LLaMA, T5, etc.)
Tokens, tokenization techniques, and embeddings explained
Utilizing pre-trained models through APIs such as OpenAI and Claude
Leveraging open-source models via Hugging Face
Prompting essentials: zero-shot and few-shot approaches
Hands-on session: crafting prompts in a Python notebook
Developing a straightforward LLM-powered application (CLI or web-based)
Understanding practical constraints: token limits, rate restrictions, and foundational reliability methods
Day 2 – RAG and Vector Search
The rationale for RAG: linking LLMs with proprietary data
Architecture of RAG systems: ingest, index, retrieve, generate phases
Preparing documents through chunking to optimize retrieval
Generating text embeddings via APIs or Hugging Face
Introduction to vector databases (e.g., Chroma, Pinecone)
Hands-on session: building a basic semantic search script
Hands-on session: implementing a document Q&A system using RAG
Scaling ingestion and embedding processes for larger data workflows
Balancing design trade-offs in RAG: chunk size, top-k selection, cost versus quality considerations
Day 3 – Workflows, Agents, and Production
Defining AI agents and identifying appropriate use cases
Introduction to LangGraph and graph-based LLM orchestration
Hands-on session: constructing a simple LangGraph workflow with integrated tools
Enhancing workflows with memory retention and multi-step reasoning
Combining RAG techniques with agentic frameworks (agentic RAG)
Monitoring and evaluating the performance of LLM and RAG systems
Deployment strategies for LLM applications: APIs, containers, and cloud services
Optimization approaches for cost efficiency and system performance
Implementing basic safety measures, guardrails, and responsible AI practices
Capstone project: delivering an end-to-end demo integrating RAG/agent features
Requirements
A solid grasp of Python programming and familiarity with APIs is required.
Target Audience:
This course is designed for organizations aiming to transition from experimental projects to fully functional LLM-powered solutions. It is ideal for software, backend, and full-stack engineers integrating LLMs into their products; data and machine learning engineers focused on RAG, embeddings, and vector search; solution and enterprise architects crafting LLM-based system designs; as well as technical product owners and engineering leaders responsible for evaluating AI use cases, associated costs, and potential risks.
Testimonials (2)
use of proper and effective prompt
Marses Pacaldo
Course - Generative AI and Prompt Engineering for Corporate Professionals
The interactive style, the exercises
