Building Autonomous Agentic AI Systems for Beginners HandsOn

Explore the foundations of Agent Key, its differences from traditional AI and AI agents, and the core building blocks shaping modern intelligent systems, plus real-world impact and limitations.

Explore how agentic AI turns large language models into autonomous agents that act toward goals with minimal supervision, interacting with the world, calling APIs, and monitoring real-time data.

The lecture contrasts generative AI, AI agents, and agentic AI—content creators, task executors, and proactive goal-oriented systems that plan and act.

Discover why organizations shift from passive generative AI to autonomous agents that reason and execute end-to-end workflows. Learn how tool use, multi-agent collaboration, and resilient automation close the execution gap.

Discover the building blocks of an agentic AI system, including perception, reasoning, action, and continuous learning through feedback loops, reinforcement and self-supervised learning.

Explore how agent AI systems automate tasks across industries, boost efficiency, and enhance decision making for developers and customers, while highlighting data quality, integration, and security limitations.

Explore the foundations of agentic ai: autonomy, reasoning, decision making. Compare with traditional ai and basic agents; outline building blocks, workflows, otc ai limitations.

Explore language models as cognitive core and ai agents as operational layer, then build a weather ai agent and learn retrieval augmented generation, embeddings, memory, and prompt engineering.

Understand large language models as the backbone of agent systems, learn how transformers enable context-aware reasoning, and see how LLMs empower AI agents to interpret goals and act intelligently.

Explore how AI agents form the action layer of agentic AI, turning language model reasoning into real-world actions, and distinguish automation from agents across reactive, proactive, and collaborative types.

Build a weather ai agent from scratch that connects language, reasoning, and real-world tools using the Gemini model, city extraction, and an open weather API to return human-like weather reports.

Learn how retrieval augmented generation (RAG) bridges static model knowledge and real-time information through embeddings, vector database searches, and context-rich responses for accurate, domain-aware AI.

Embeddings and vector representations power agentic ai by turning text into vectors in a vector space to measure meaning, enabling memory and semantic search with Pinecone and Milvus.

Explore how memory systems empower agentic AI to learn, adapt, and personalize interactions. Distinguish short term, long term, episodic, semantic, and procedural memory, plus context windows and vector-based retrieval.

Craft prompts and orchestrate tasks to guide llms through reliable, multi-step workflows using zero-shot, few-shot, chain of thought prompting, and structured prompting.

Understand the end-to-end agentic ai workflow from user query to task execution, with the llm as the cognitive core and the agent orchestrating actions, retrieval, memory, prompts, and task orchestration.

Demonstrates an ai powered AWS instance launcher using a Flask interface, where a user prompt triggers Gemini language model to create an EC2 instance via Boto3.

Build and apply AI agents from scratch using large language models, rag embeddings, vector representations, and memory systems, then orchestrate prompts and deploy real-world automation via the AWS Instance Launcher.

Explore modern frameworks and platforms to build functional agent ai systems, and gain hands-on experience with small agents, a weather agent, and sql workflows.

Discover the technology stack for building autonomous ai systems, from foundation models through supporting tools to frameworks and their code, low-code, and no-code variants.

Learn to build an ai agent with the small agents framework from Hugging Face, install the library, and run a simple sum from 1 to 10 with token authentication.

Demonstrates building a weather agent with smolagents, an open source Python library, in lines of code using a weather API and web search tool to fetch the city's temperature.

Demonstrate building an sql query agent that turns natural language into sql and runs safe queries in an in-memory sqlite database.

Shows how to build a low-code, no-code ai agent with the Natan platform, linking trigger, memory, and http request tools to fetch the latest tech news.

Develop agent systems across frameworks and platforms, demonstrate weather agent and SQL capabilities with Agent Key, then automate workflows with a low-code, no-code agent using natin for real-world applications.

Explore the model context protocol (MCP) to enable secure agent communication with external tools and environments. Build an MCP server and an agent, then demo Docker MCP for real-world infrastructure.

Discover how the Model Context Protocol enables llms to connect with tools and data sources securely and uniformly. MCP acts as a universal bridge, turning llms into intelligent agents.

Discover how MCP orchestrates AI actions through the host, client, and server, using the MCP protocol and JSON RPC 2.0, with shared context for smarter tool use.

Build a minimal MCP compatible server that supports discovery and a get weather tool, using fast api and json-rpc, with an in-memory weather dataset and a simple client script.

Create a lightweight agent on an MCP server that interprets queries using an LLM. Extract the city, call the MCP weather tool, and deliver a natural language weather update.

Demonstrates connecting a cloud desktop client to a Docker MCP server to manage containers with an agentic AI, including listing, creating, and deleting containers via the MCP toolkit.

Master MCP by connecting agents to external tools and environments, building a complete MCP server, and running a Docker MCP setup for containerized, production-ready integration.

Explore multi-agent systems where specialized agents collaborate, communicate two-way, and split tasks under centralized orchestration for efficient, scalable, fault-tolerant automated workflows.

Explore the core components of multi-agent systems, including agents, environment, interactions, and communication, and compare reactive, deliberative, and hybrid architectures for scalable collaboration.

Explore frameworks for real-world multi-agent systems, enabling communication, coordination, scheduling, and decision making. Compare Jade, paid, Mesa, Lang Chain, Land Graph, Autogen, and Cru AI as the practical, production-ready choice.

Install Crew AI on your machine, verify the setup, and review its documentation to learn about role-based agents, designated tools, processes, tasks, and multi-agent collaboration.

Build a simple multi-agent system with crew AI in python, powered by Gemini, using research and writer agents and the server dev tool to generate a markdown report.

Explore a three-agent, Crew AI powered multi-agent system that analyzes customer support data with a custom data fetcher, identifies bottlenecks, and produces an executive report.

Demonstrates building a three-agent, weekly newsletter workflow in n8n using Tavoli for research, a planning and section writing process, an editor step, and automatic draft creation in Gmail.

Build a real-world AWS monitoring multi-agent system that fetches EC2 data with Boto3, analyzes it with the Gemini LM, and produces a markdown report on running instances and monthly cost.

Demonstrate a cross-cloud multi-agent system with AWS and Google Cloud, where agents fetch and summarize data, the analyst creates a consolidated report, and the mailer delivers it by email.

Explore multi-agent systems, their architecture, and frameworks, then demonstrate building simple and advanced agents, no-code workflows, and real-world applications like AWS monitoring and cross-cloud coordination.

Explore retrieval augmented generation (rag) and its role in enhancing AI agents with real-time retrieval, reasoning, and multi-agent rag for contextual insights like course intelligence and project feasibility analysis.

Integrate rag into ai agents by embedding a retrieval layer into the decision loop, grounding actions in up-to-date external documents via a vector database of embeddings and similarity search.

Build a rag-based Q&A agent with a Streamlit UI in Python, uploading a file and chatting with its content using Gemini and Google generative AI.

Demonstrates building an agentic rag with crew AI, featuring a retriever and a responder, both powered by a Gemini LLM and a knowledge base.

Demonstrate building an agent rack system that uses a large language model with retrieval augmented generation (RAG) on a course knowledge base of PDFs to fetch precise information.

Demonstrates a project feasibility agent that analyzes proposals against an organizational knowledge base using rag-based vector database and Gemini Elm, delivering a report with timeline, budget, risks, and tender status.

Enhance agents with retrieval augmented generation, integrating retrieval, reasoning, and decision making for responses, then build a rag-based q&a agent and apply it to course intelligence and project feasibility analysis.

Turn agents from task executors into self-improving, production ready systems by exploring reflection, self-improvement, deployment principles, and guardrails for safe, reliable operation.

Explore how agent ai systems reflect on their reasoning and use observation, evaluation, and revision within a three-level reflection loop—immediate, episodic, and long-term—to drive self-improvement, memory, and safer, adaptive performance.

Deploying agents moves from build to operation by packaging, hosting, and scaling modular layers from model to interface. Leverage containers and cloud platforms with observability and security to stay reliable.

Explore guardrails and safety for autonomous agents, including input, model, and output safeguards, plus ethical alignment, privacy, and transparency to build trusted, responsible AI systems.

Explore advanced concepts that turn agents into autonomous systems, using reflection to analyze actions and improve performance. Learn to deploy agents with guardrails and safety for ethical, secure, controlled operation.

Explore agentic design patterns to structure agentic ai systems, learn how to select patterns by task requirements and constraints, and compare single and multi-agent architectures.

Explore agentic design patterns as reusable blueprints guiding multi-agent workflows and tool integrations. Structure complex systems for scalability and reliability, and manage task decomposition, delegation, and coordination through proven patterns.

Study single agent design patterns, where one ai agent handles tasks end-to-end using a model, a system prompt, and tools. Understand workflow and how this foundation leads to multi-agent architectures.

explore advanced multi-agent design patterns: iterative refinement, coordinator, hierarchical decomposition, and swarm. understand how each pattern affects quality, latency, and cost, and how they balance flexibility with reliability.

Build and deploy autonomous agent AI systems through hands-on practice, mastering agent HCI, memory embeddings, rag and prompt orchestration, and multi-agent collaboration.