Build AI Agents: CrewAI, LangGraph, OpenAI SDK, Autogen, MCP

Present the foundations of artificial intelligence, large language models, and APIs, explaining input-to-model processing and outputs, training data quality, and capabilities like generating text and code.

Explore the architecture and file structure for a research assistant project, wiring config, llm client, and prompts, and implement a gradio user interface to search and summarize topics.

Build an ai email generator that outputs tailored emails by selecting tone and purpose. Explore real-world uses in support, sales, hr, and marketing, and show how prompts shape the emails.

Builds an ai email generator by outlining a project structure, configuring the LLM client, crafting prompts for tone, purpose, and context, and integrating Gradio UI.

Build your first ai chatbot using a gradio interface. Format chat history in markdown, store history, and manage system, user, and assistant roles, and connect to the OpenAI GPT model.

Explore how tokens and probability drive LLM thinking, including input and output tokens, prompts, context windows, costs, and the risk of hallucinations.

Explore what LLMs can do—code, summarize, explain concepts, and generate emails—while recognizing limits like hallucination, weak math reliability, prompt sensitivity, and context windows; validate outputs.

Explore cost and latency basics for AI apps, including input and output tokens, prompt size, and model choice to optimize expenses and response times.

Explore building a prompt playground to compare prompts and outputs, select prompt styles (simple, beginner, friendly, expert, step-by-step), and learn how prompt engineering shapes results.

Explore a prompt playground that compares prompts and outputs by implementing a structured project with a main UI and diverse prompt styles; learn how prompts shape AI responses and cost.

Generate structured AI outputs from unstructured input by converting customer issues into a table-ready format with fields like customer, issue type, product, and summary, enabling analytics on payments and subscriptions.

Build a json output formatter that converts unstructured text into a json with fields like customer name, issue type, product, priority, and summary, using prompts to enable api integration.

Build an email intent classifier that categorizes messages into support, sales, and spam, routes them to the right teams, and outputs a valid JSON with priority.

Design and test an AI text improver that enhances grammar, clarity, tone, and professionalism by using prompts, a system prompt, and a Gradio UI to input text and select tone.

Build an ai data extractor that converts text from PDFs and documents into a structured json, enabling automated document processing workflows for invoices, contracts, and resumes.

Identify and mitigate failure modes in AI agents and LLM systems, including hallucinations, invalid outputs, tool and prompt failures, latency, cost, security risks, and multi-agent coordination.

Explore building an ai sql assistant that generates, explains, optimizes, and debugs sql across multiple databases like Postgres and MySQL.

Explore a multi-step task agent that plans, executes, reviews, and refines tasks iteratively to mirror human workflows and decompose work into stages for robust reasoning.

Learn to build a multi-step task agent that plans, executes, and refines tasks using planning, execution, and refinement prompts, wired into a main method and gradio UI.

Demonstrate building a research copilot with the OpenAI software development kit that uses a web search tool to generate a structured report through multi-step reasoning.

Builds an enterprise AI workflow that transforms meeting notes into action items using the open AI agents SDK, identifying owners, responsibilities, and next steps.

Explore crew AI fundamentals, including role-based agents, tasks, and crew collaboration for multi-agent orchestration. Compare single-agent limitations and the benefits of specialized, YAML-configured agents for scalable, modular workflows.

Create an AI hiring assistant that screens resumes, analyzes skill gaps, and designs interview questions using a multi-agent workflow, with a two-input UI for resume and job description.

Explore LandGraph's graph-based, stateful customer support workflow to classify tickets, route by category, apply business rules, and escalate when needed, all while maintaining shared state across nodes.

Learn to build a human-in-the-loop approval workflow that automatically auto-approves low-risk tickets and routes others for human approval, with risk analysis and interactive UI.

Build collaborative AI code review agents with Autogen, assigning developer, code reviewer, security, and testing roles, each powered by different LLM models in a managed group-chat workflow.

Build problem-solving agents with Autogen to tackle complex multi-step tasks through a collaborative multi-agent workflow, including a problem analyst, solution strategist, critic, and final answer agents.

MCP, the model context protocol, standardizes connecting AI systems to tools, data, and external services, like a USB for AI. It defines tools, resources, and client-server interactions.

Understand guardrails as rules, validations, and controls that constrain ai behavior, preventing unsafe outputs, data leakage, and hallucinations, with input, output, tool, policy, structured, security, and human-in-the-loop guardrails.

Build an enterprise AI internal data assistant with MCP that converts natural language questions into safe, read-only SQL queries against a SQLite database, enforces guardrails, and summarizes results.

Builds a knowledge base agent from uploaded PDFs using MCP, guardrails, and prompts to answer questions strictly from the document.

Build a trading simulation agent that fetches Yahoo Finance data via YFinance, applies guardrails to validate risk, and uses an LLM to generate cautious buy, hold, or sell recommendations.