History of AI: from Turing and expert systems to deep learning and generative AI · Types of AI: narrow AI, generative AI, multimodal AI, and AGI — current state · Machine learning: supervised, unsupervised, and reinforcement learning fundamentals · Neural networks: architecture, activation functions, and backpropagation (conceptual) · Key milestones: AlexNet, AlphaGo, GPT-3, ChatGPT, GPT-4, Claude 3, Gemini, Llama 3 · AI ecosystem: OpenAI, Anthropic, Google DeepMind, Meta AI, Mistral AI · AI limitations: hallucinations, knowledge cutoffs, context windows, and bias

How LLMs work: tokenisation, embeddings, transformers, and self-attention · Training process: pre-training, instruction fine-tuning, RLHF, and Constitutional AI · Context windows: GPT-4o (128k), Claude 3.5 (200k), Gemini 1.5 Pro (1M tokens) · Model families: GPT-4o, Claude 3.5, Gemini 1.5 Pro, Llama 3.1, Mistral, Phi-3 · Multimodal models: text, image, audio, video, and document inputs · Fine-tuning vs prompting vs RAG — cost, performance, and selection criteria · Open-source models: running Llama and Mistral locally with Ollama · Evaluating LLM outputs: BLEU, ROUGE, human evaluation, LLM-as-judge

What is prompt engineering? The skill, craft, and science · Zero-shot prompting: direct instruction without examples · Few-shot prompting: providing examples to guide model behaviour · Chain-of-thought (CoT): step-by-step reasoning — standard and zero-shot CoT · Role and persona prompting: system message design and voice control · Output format control: JSON, markdown tables, XML, and structured data · Iterative prompt refinement: diagnosing failures and systematic A/B testing · Hands-on lab: 20 real prompt engineering challenges across writing, coding, and analysis

Tree-of-thought (ToT): exploring multiple reasoning paths and self-evaluation · ReAct prompting: reasoning and acting — thought, action, observation loops · Self-consistency: sampling multiple chains and selecting majority answer · Prompt injection attacks: direct/indirect injection, jailbreaking, prompt leaking · Defensive prompting: input sanitisation, guardrails, robust system prompt design · Agentic AI: agents, tool use, function calling, and multi-step reasoning · Multi-agent systems: orchestration, routing, handoffs, and collaboration · LangGraph and CrewAI: building stateful multi-agent workflows · Hands-on lab: build a ReAct agent that uses tools to answer complex questions

Retrieval-Augmented Generation (RAG): architecture, benefits, limitations · Embeddings: vector representations and semantic meaning · Vector databases: Pinecone, Weaviate, ChromaDB, Qdrant — similarity search · Building a RAG pipeline: chunking, embedding, indexing, retrieval, generation · Advanced RAG: query rewriting, HyDE, re-ranking, and hybrid search · LangChain and LlamaIndex: orchestrating LLM chains and document pipelines · OpenAI and Anthropic APIs: authentication, parameters, streaming, best practices · Deploying AI apps: Streamlit, Gradio, FastAPI, and Hugging Face Spaces · Hands-on lab: build a full RAG-powered document Q&A system with LangChain

AI ethics frameworks: beneficence, non-maleficence, autonomy, and justice · Bias in AI: dataset bias, algorithmic bias, detection, and mitigation · Transparency: LIME, SHAP, and communicating AI decisions to stakeholders · AI safety: alignment, reward hacking, and Constitutional AI · Privacy in AI: data minimisation, differential privacy, and GDPR compliance · EU AI Act (2024): risk categories, prohibited uses, and high-risk obligations · Red-teaming: adversarial testing, failure mode discovery, responsible disclosure · Capstone: design, build, and present an AI-powered application for a real problem · Ethical reflection report: document risks, mitigations, and societal impact

Master prompt engineering from zero-shot basics to advanced agentic AI workflows · Build production AI applications using LangChain, RAG, and the major AI APIs · Evaluate and select the right LLM for any business or research use case · Design responsible AI systems compliant with the EU AI Act and global ethical standards · Achieve a credential aligned to OpenAI, Anthropic, Google AI, and EU AI Act standards

Complete all 6 modules, pass a 60-question proctored examination (minimum 75%), submit a deployed AI-powered application with GitHub documentation, and write a 1,000-word ethical AI reflection report.

AI Engineer, Prompt Engineer, ML Engineer, AI Product Manager, AI Consultant, LLM Application Developer, AI Safety Researcher. Average starting salary: $80,000–$140,000 USD.