Agentic AI for Beginner: Build AI Agents From Scratch

Design a multi-agent system by defining each agent’s role, delivery, and starting context, and ensure collaboration to achieve the project goal; illustrated with a chef and nutritionist example.

Define and build a first multi-agent system with two agents, chef and nutritionist, and define tasks to cook and critically analyze recipes for diabetes.

Build a crew of artificial intelligence agents, like a chef and a nutritionist, to generate five spinach recipes, delegate tasks, and optimize calorie counts for a diabetes-friendly final dish.

Organize a customer service multi-agent system where a customer care specialist drafts a response and a quality check agent reviews it for accuracy in CRIU AI.

Explore building a multi-agent system with Microsoft Autogen, a framework for orchestrating LLM workflows, using grok cloud api key and .env setup to run a basic conversable agent.

Explore building a four-agent investment analysis team using fidata, assigning market sentiment, financial data, financial expert, and investment strategy roles, with memory, tools, and reasoning.

Build an agentic four-agent investment analysis team to fetch latest data, analyze sentiment and fundamentals, and output a data-driven buy, hold, or sell recommendation with transparent tables.

Lead a multi-agentic investment analysis team that uses sentiment analysis, financial data, and tool integrations to produce data-backed buy, hold, or sell recommendations for AMD and TSM.

Build a low-code AI agent with Langflow by adding tools, enabling web search via Tavaili, and running the agent from backend with a Python API for front-end and backend deployment.

Explore the basics of AWS, including services like S3, DynamoDB, SageMaker, and Bedrock, and learn to set up IAM users, access keys, and the AWS CLI for hands-on access.

Discover AWS Bedrock for building applications with large language models, using the model catalog and playground, enabling model access, and calling models via boto3 or the Bedrock runtime.

Learn to build and test your first Bedrock agent with the builder tool, configure a service role, select Titan Text G1, set instructions, prepare the agent, and create an alias.

Learn to call an agent from Python with boto3, uuid, and a role ARN, and resolve access denied errors by editing the trust relationship for STS assume role.

Design action groups in AWS Bedrock to enable a smarter agent, with one group recommending courses by years of experience and a second explaining why, using lambda functions in Python.

Analyze agent observability through the bedrock reasoning trace to see how action groups and function calls drive course recommendations for AI and deep learning.

Build a user interface to interact with an AWS-hosted agent using Streamlit and Python, with setup in Visual Studio Code, a requirements.txt, and a basic Streamlit app.

Create a chat input and send button to capture user messages, append them to chat history, and invoke the agent within the same session, parsing event stream chunks and traces.

Perform system validation by parsing and validating agent outputs, store chat history, and implement try-catch error handling, then rerun the Streamlit UI to demonstrate a new alias and final demo.

Celebrate completing the course and learn how ratings and feedback motivate the instructor, while you explore other courses and stay connected via the platforms mentioned.

Explore LangFuse for AI observability and tracing to debug LLM applications. Set up a project, API keys, and basic Colab tracing to inspect inputs, outputs, latency, and costs.

Discover an hr onboarding assistant built with three agents—parser, QA, and task—and a central orchestrator that coordinates calls, enabling agent observability and tracing with Langfuse.

Explore how an HR onboarding assistant uses the OpenAI Agent SDK and Langfuse observability to parse onboarding documents, answer policy questions, and manage tasks across three agents with an orchestrator.

Implement a custom Langfuse tracing processor to bridge the OpenAI SDK with Langfuse, wiring app.py and the processor into the agent tracing flow for agents, responses, handoffs, and functions.

Explore implementing a two-layer tracing system with a custom LangFuse tracing processor, capturing data at the span level (agent, response, generation, function, handoff) and linking it to the app.

Build an AI agent in Make AI without code, connect OpenAI, and enhance it with FAQs, live updates, and Google document content through scenarios and tools.

Learn to diagnose Python package issues by using a versioned requirements.txt, install with pip, and manage virtual environments and kernels, plus troubleshoot API key errors.