🤖 AI Academy Syllabus

• 01 Basics What is an LLM and Tokens This module establishes the foundational mental model for Generative AI. We transition from viewing AI as a "magic box" to understanding it as a high-dimensional probabilistic engine. We cover the transition from characters to sub-word tokens and why this architectural choice dictates the intelligence, cost, and latency of systems like Gemini or GPT-4. ▶️ Live Output

• 02 Basics Prompting and Temperature Pending... ▶️ Live Output

• 03 Milestone Interactive Prompt Playground In this milestone, we transition from static scripts to a dynamic, interactive playground. You will learn how to build a tool that allows you to experiment with System Prompts, hyper-parameters (like Temperature and Top-P), and different Prompt Engineering techniques in real-time. This environment is essential for "vibe-checking" models and fine-tuning instructions before deploying them into production workflows. ▶️ Live Output

• 04 API Connecting Python to Gemini > Learn how to give your Python scripts "eyes and a brain" by connecting them to Google's most powerful AI model, Gemini. ▶️ Live Output

• 05 API Structuring JSON Responses When an API talks to an app, it shouldn't just "shout" information. It needs to organize that information into a specific format called JSON (JavaScript Object Notation). In this lesson, we learn how to structure these responses so they are predictable, clean, and easy for any developer to use. ▶️ Live Output

• 06 API Security Handling Secrets In this lesson, we learn the golden rule of API development: Never put your passwords or API keys directly into your code. We will explore how professional developers use environment variables and "vaults" to keep sensitive information safe from hackers and accidental leaks. ▶️ Live Output

• 07 Basics Understanding Context Windows Learn about the "memory limit" of Large Language Models (LLMs). We explore the Context Window—the specific amount of information an AI can process in a single go—using the analogy of a workspace and a real-world look at how Google handles massive data. ▶️ Live Output

• 08 Milestone The Smart Summarizer App In this milestone, you will build an AI-powered application that takes long, complex articles and condenses them into short, digestible bullet points. This tool solves "Information Overload" by acting as a personal filter for your brain. ▶️ Live Output

• 09 Logic Reasoning and Chain of Thought In this lesson, you will learn how to guide an AI to solve complex problems by breaking them down into logical steps. Instead of jumping straight to a conclusion—which often leads to errors—Logic and Chain-of-Thought (CoT) techniques force the model to "show its work," leading to higher accuracy and more reliable reasoning. ▶️ Live Output

• 10 Logic Self Correction Loops In AI and software engineering, a Self-Correction Loop is a process where a system reviews its own output, identifies errors against a set of rules, and tries again. The "10" represents a threshold—allowing the system up to 10 attempts to perfect its logic before finalizing the result. ▶️ Live Output

• 11 Memory What are Embeddings In this lesson, we explore the concept of Embeddings—the secret sauce that allows Artificial Intelligence to understand the "meaning" of words rather than just the characters. You will learn how computers turn concepts into numbers and how they use those numbers to find relationships between different pieces of data. ▶️ Live Output

• 12 Memory Vector Databases Explained A Vector Database is a specialized storage system designed to handle "embeddings"—numerical representations of data (text, images, or audio) that capture their meaning. Unlike traditional databases that look for exact keyword matches, vector databases look for "similarity," allowing AI to have a long-term memory. ▶️ Live Output

• 13 Memory Similarity Search Lab This lab teaches you how AI systems "remember" and find information. Instead of searching for exact words (like a Ctrl+F search), Similarity Search allows AI to find information based on meaning. You will learn how to turn text into mathematical coordinates (Embeddings) and find the "closest" answers in a Vector Database. ▶️ Live Output

• 14 RAG Chat with your PDFs Learn how to build a Retrieval-Augmented Generation (RAG) system. This lesson covers the concept of giving an AI a "brain upgrade" using your own documents, using a simple open-book exam analogy and a real-world tech example. ▶️ Live Output

• 15 RAG Web Search Integration Standard RAG (Retrieval-Augmented Generation) is limited to the documents you give it. This lesson teaches you how to integrate real-time Web Search so your AI can answer questions about breaking news, current events, and live data using tools like LangChain and Tavily. ▶️ Live Output

• 16 RAG Hybrid Search Techniques Hybrid Search is a technique in Retrieval-Augmented Generation (RAG) that combines Keyword Search (looking for exact words) and Semantic Search (looking for meaning). This ensures that if a user searches for "The Dark Knight," the system finds the exact title AND understands they are looking for "Batman movies." ▶️ Live Output

• 17 Memory Long Term Chat History Standard AI chatbots have "goldfish memory"—they forget everything once you refresh the page. This lesson teaches you how to give your AI a "permanent diary" using Long-Term Memory, allowing it to remember user preferences, past names, and old conversations across different sessions. ▶️ Live Output

• 18 Milestone Personal Knowledge Brain Learn how to build a "Second Brain"—a digital system that uses Artificial Intelligence to store, organize, and retrieve every piece of information you've ever learned, making you "forget-proof." ▶️ Live Output

• 19 Vision Analyzing Images and Charts In this lesson, you will learn how "Vision" models (Multimodal AI) bridge the gap between pixels and language. You'll discover how AI can describe photos, interpret complex business charts, and even troubleshoot code from a screenshot. ▶️ Live Output

• 20 Vision Video Understanding AI Video Understanding AI (also known as Vision-Video Understanding) is the evolution of computer vision. While traditional AI can identify an object in a picture, Video Understanding AI analyzes movement, intent, and sequences across time. This lesson covers how AI evolves from seeing "frames" to understanding "stories." ▶️ Live Output

• 21 Agents Function Calling Basics In this lesson, you will learn the "bridge" between thinking and doing. While Large Language Models (LLMs) are incredibly smart at talking, they are "locked in a room" with no access to the outside world. Function Calling is the mechanism that gives the LLM "hands," allowing it to interact with databases, APIs, and software tools to perform real-world tasks. ▶️ Live Output

• 22 Agents Tools Calculators and APIs Large Language Models (LLMs) are like brilliant professors who have read every book but are locked in a room without a clock, a calculator, or internet access. "Agents" are the framework that allows these professors to use tools—like calculators for math and APIs for real-time data—to solve complex, real-world tasks. ▶️ Live Output

• 23 Agents Building a Web Researcher Learn how to build an automated research team using AI agents. This lesson covers how multiple specialized agents (Searcher and Writer) collaborate to browse the internet, synthesize information, and create comprehensive reports using CrewAI and Streamlit. ▶️ Live Output

• 24 Agents Multi Task Orchestrator The Multi-Task Orchestrator is a design pattern in AI where a central "Master Agent" receives a complex goal, breaks it down into smaller sub-tasks, and delegates those tasks to specialized "Worker Agents." Instead of one AI trying to do everything poorly, this system uses multiple specialized AIs to do specific things perfectly. ▶️ Live Output

• 25 Agents AI Coding Assistant Learn how to move from a single AI chatbot to a massive "Multi-Agent System" (MAS). Instead of one AI trying to do everything, we use 25 specialized AI agents—each an expert in a tiny niche (like Security, Testing, or Documentation)—to build complex software projects collaboratively. ▶️ Live Output

• 26 Swarms Manager and Worker Agents In this lesson, you will learn how to build a sophisticated AI "Company" using the Swarms framework. We will move beyond single-agent bots to a hierarchical system where a Manager Agent directs multiple specialized Worker Agents to complete complex, multi-step projects. ▶️ Live Output

• 27 Swarms Collaborative Content Team Imagine if you didn't just have one AI writing for you, but an entire team of specialized AI experts—a researcher, a writer, an editor, and an SEO specialist—all talking to each other to produce a perfect article. This is the "Collaborative Content Team" pattern (often referred to as Swarm #27 in multi-agent frameworks). It moves away from "one-shot" prompting to a structured pipeline where agents collaborate to ensure high-quality output. ▶️ Live Output

• 28 Advanced Fine Tuning Concepts Basic fine-tuning is like teaching a student to memorize a textbook. Advanced Fine-Tuning is like training a specialist to perform surgery using minimal resources, high precision, and human-like judgment. This lesson covers the shift from "retraining everything" to "efficiently adapting" Large Language Models (LLMs) using techniques like LoRA, QLoRA, and RLHF. ▶️ Live Output

• 29 Advanced AI Ethics and Safety As AI systems become more powerful, we must ensure they are fair, transparent, and safe. This lesson covers how to move from "it works" to "it works responsibly," focusing on bias detection, explainability, and safety guardrails. ▶️ Live Output

• 30 Final Capstone The Autonomous Company The ultimate goal of AI integration is the "Autonomous Company"—a business entity where AI agents handle the strategy, operations, marketing, and customer service with minimal human oversight. This capstone explores how to orchestrate multiple AI components into a self-sustaining loop. ▶️ Live Output