The Agentic AI Engineering Masterclass 2026

Build your first ai agent with the OpenAI agents SDK in a Jupyter notebook, set up the environment, create a simple fact checker, and trace its actions with a runner.

Set up the OpenAI API key and environment by installing the OpenAI agents library, creating and loading a dot env key, and running a quick sanity check to verify access.

Build and run your first ai agent using the OpenAI agents library, set up the api client and brain, design a clear system prompt, and run with a runner.

Experiment with the OpenAI agents SDK by changing the statement variable, test the agent with new input like the tallest mountain, and explore different models such as GPT five mini.

Learn to build and test an ai agent by selecting models—from gpt five mini to nano and frontier open-source options—and running a new input with a fast checker agent.

Demonstrates testing an AI agent's memory by asking what was discussed previously, explains traces and tokenization in the OpenAI API, and highlights monitoring tools and token optimization.

build tweetbot, ai agent using OpenAI agents sdk, to generate a tweet on the future of renewable energy with one emoji and one hashtag, tracing behavior and token usage.

Develop a tweet generator ai agent by issuing role-based instructions to produce a tweet under 280 characters with an emoji and a hashtag on a topic using GPT five mini.

upgrade a stateless ai agent by adding memory with a SQLite session, building a market research assistant that analyzes companies and performs multi-turn interactions using the OpenAI agents sdk.

Learn to build a stateless AI agent with no memory by setting up the environment, configuring the OpenAI API, and running a market researcher agent through basic Q&A.

Add memory to your ai agent with a sqlite session to store conversation history, enabling context, memory embedding, and serverless, open-source memory management for improved responses.

Build a travel planner ai agent with memory using OpenAI's Agents SDK that always suggests one sunny, warm weekend getaway destination, stores conversations in SQLite, and tests from Toronto.

Build an ai agent with memory to plan weekend getaways, choosing sunny destinations within five hours of Toronto and answering visa requirements for Canadians.

Build ai agents that leverage Tavli search and OpenAI tools. Set up OpenAI api keys and wrap a custom tool for end-to-end workflows.

Set up the OpenAI API and Tavileh integration, create and load API keys into the environment, and test key access to enable real time web search tools for AI agents.

Define a Tavelli search function and expose it as an AI agent tool using typed dictionaries, type hints, and the function tool decorator.

Modify the search function to return only titles, exclude content, and extend the max results from two to three to practice AI workflow automation skills.

Solve and review the practice opportunity solution, modify the Tavily search function to return only titles with a default max of three, and prepare for invocation in upcoming tasks.

Build and run an AI agent with a decorated search tool and memory via SQLite, configured as a live market researcher, and use traces to verify real-time web search results.

Practice opportunity trains testing AI agents with tools and memory by using a live researcher agent to extract and summarize vehicle features, test memory, and perform searches.

Practice opportunity solution teaches you to test ai agents with tools and memory by running a live researcher agent, posing questions, and summarizing results.

Leverage built-in OpenAI tools, including the code interpreter, for real-time web/file searches, vector stores, and retrieval augmented generation to analyze data and simulate tariff-driven prices.

Build and orchestrate a team of specialized ai agents—researcher, analyst, and writer—to achieve goals through modular roles, shared outputs, and a guiding manager function.

Set up OpenAI API keys and required tools, install the OpenAI agent 0.2.2, and define the tavli search function as a tool with a decorator.

Define two AI agents in the OpenAI Agents SDK—researcher and analyst—specifying data types, inputs, and outputs to enable search-driven research summaries and analyst insights.

Run the researcher and analyst agents to analyze trends in batteries for electric vehicles. Form a research question, fetch findings, and pass summaries between agents to reveal trends and risks.

Practice opportunity solution shows how to run both ai agents, pass the researcher agent’s electric vehicle battery summary to the analyst agent, and produce a concise final analysis.

Define a writer agent that outputs an executive summary, a detailed markdown report, and 3–5 follow-up questions, using pedantic data validation and clear input-output structure.

Rewrite the writer agent's instructions to generate all outputs in French, producing a concise French report, and prepare a manager function to integrate these components for subsequent testing.

Learn to update writer agent instructions, enforce language and emphasis cues, prevent hallucinations, ground outputs in documents, and generate French reports using shift-enter to define the agent.

Build and orchestrate a full artificial intelligence pipeline by running researcher, analyst, and writer agents with session memory to generate a traceable, complete report.

Build an AI agents pipeline: French report writer, creative director to brainstorm ideas, strategist to select top two, and copywriter to draft tweets for a Bali eco-friendly water bottle launch.

Learn to build and orchestrate multi-agent pipelines for creative advertising, using a creative director, strategist, and copywriter to generate top campaigns and engaging tweets in multiple languages.

Build a multi-agent system with planner, writer, search, and fundamental analyst agents, powered by OpenAI API, using guardrails, handoffs, and tools to produce executive reports.

Install OpenAI agents v0.2.2, load libraries, and enable guardrails with API keys. Create a search tool for AI agents using a query and max results, plus main and nano models.

Define and test a planner ai agent with guardrails that splits queries into three web searches with reasons and queries, plus a politics guardrail.

Design and test a defense guardrail for AI agents to block defense and military topics, using examples like Lockheed Martin drones to verify detection and triggering.

Explore how to implement a defense guardrail in an ai planner agent, including triggering a tripwire for defense topics, politics, and government policy, with test runs and reasoning.

Define a team of ai agents, a searcher and a fundamentals analyst, as tools for another agent, enabling up-to-date web searches and concise fundamentals analysis.

Design an ai writer agent that uses search and fundamentals tools as reusable companions to craft a comprehensive markdown report with an executive summary and follow-up questions.

Explore multi-agent traces on the OpenAI API platform, analyzing tool usage and counts, explain the trace hierarchy and workflow, and perform a quick sanity check for AI agent workflows.

Explore multi-agent traces on the OpenAI API platform, learning how a writer agent delegates to a searcher and fundamentals analyst, with nested tool calls and trace hierarchy.

Clone the planner, implement handoffs to the writer using the OpenAI agents SDK, and enforce planner-to-writer input with guardrails.

Instantiate a sentiment analysis AI agent, update its instructions to focus on online sentiment for companies, and test it with a Tesla input alongside existing search and fundamentals agents.

Build AI agent workflows with a sentiment tool, planner, and writer agent to generate executive summaries and market sentiment reports. Master guardrails, handoffs, and tools integration for insightful analyses.

Build an ai agent team with AutoGen, including chief marketing officer, a marketing manager, and social media strategist, powered by ai, Gemini, and Anthropic, with a human in the loop.

Explore Autogen to orchestrate multiple AI agents that communicate and collaborate using LLMs, tools, and human-in-the-loop capabilities. Learn group chat patterns with a group chat manager.

Design a 2 to 3 distinct roles for an AI agents team to plan a vacation, detailing each agent's specialization and interactions, and craft messages to guide their behavior.

Design a three-agent AI team to plan a vacation: destination research agent matches preferences with destinations, logistics and booking agent handles flights and hotels, and local experience agent builds itineraries.

Build your first AI agents with Autogen and OpenAI, configuring prompts and lm settings. Explore planner and doer roles, and agent communication for a marketing workflow.

Create a third AI agent for the social media strategist in AutoGen to launch a coordinated campaign, defining its system message with the same OpenAI configuration and no human input.

Discover how to build AI agents in AutoGen by configuring a social media agent with a system prompt, platform choices, posting times, and engagement strategies.

Initiate and evaluate two OpenAI agents conversing to generate a concise campaign concept for a sustainable shoe line, including audience, channels, and kpis, then review and export the chat history.

Modify the max turn parameter in the initial chat function to six to reduce communication cost, and monitor whether the conversation extends from four to six turns.

Modify initiate_chat() parameters to set max turns to six, observe longer conversations as multiple models enhance results, with brand marketer OpenAI and Chief Marketing Officer feedback, while noting token costs.

Configure multi-model AI agents in Autogen by integrating Google Gemini and OpenAI GPT-4o, create separate LM configs, and rename agents to differentiate providers.

Configure an AI agent using Anthropic's Claude via cloud API for a social media strategist, with external research and guidance from cloud API documentation and generative AI tools.

Configure an AI social media strategist to use Anthropic's Claude via API, set the API key and cloud model, and tune the temperature and timeout for cloud LLM deployment.

Trigger multi-model ai agents conversations with a multi-model team to launch a sustainable shoe campaign, refine kpi targets, and set traffic and follower goals across social channels over four turns.

Adjust the temperature of Gemini's chief marketing officer and brand marketer to 1.6 to explore higher creativity, showing how low temperatures yield deterministic results and 1.6 produces erratic, drunk-like behavior.

Explore how adjusting creativity levels via temperature settings affects AI models across Google Gemini and OpenAI. See how high temperature can produce erratic outputs and highlight performance impacts.

Add a human as a proxy user and orchestrate a four-participant group chat with a CMO, a brand marketer, and a group chat manager AI.

Add a cloud-powered social media strategist agent to the AI agent group chat, involve the user with the CMO and brand marketer, then rerun the chat to practice team integration.

Add the social media agent to the chat with the user proxy, cmo, and brand marketer. Configure ten runs and a 0.6 temperature to build a sustainability-focused social media strategy.

Learn how Autogen enables multi-agent systems with multiple LLMs, enabling collaboration and human oversight. Recap building AI agents, multimodal configurations with Gemini and OpenAI, and group chat management.

Build an AI agent that books hotels and flights, conducts web searches, checks weather, and researches destinations using Lang Graph, with Amadeus and OpenAI API tools and a Gradio UI.

Build stateful multi-agent workflows in Lang Graph by defining nodes, edges, and a graph state. Augment LLMs with memory, tools, and conditional edges, while ensuring human-in-the-loop control and streaming.

Build your first agentic workflow in land graph by defining a simple summarization agent that uses a single LLM step to summarize input text and store the result in state.

Build your first agentic workflow in Lang Graph by defining a state graph, adding a summarize step node, and wiring an entry point to convert input into a summary.

Test and refine a summarization AI agent within a LangGraph workflow by changing the input text, running and rerunning the process, and assessing the output.

Test a summarization AI agent in LangGraph by replacing sample text, compiling the workflow, and invoking it to generate a concise one- or two-line explanation of how the internet works.

Create a LangGraph workflow that summarizes input text and translates the summary from English to Spanish using a GPT 4.0 mini model, with a two state summarize and translate flow.

Define a new sentiment step using GPT-4 mini to classify translated text as positive, negative, or neutral and update the agent state with an analyze sentiment node in Landgraaf.

Add a sentiment analysis node to an AI agentic workflow, translate the summary into Spanish, and evaluate sentiment using a structured graph of nodes and edges.

Learn to bind a Tavileh web search tool to your LM in the Lang Graph framework and add a conditional edge to trigger tool calls.

learn to build an ai workflow with one tool and conditional edges, binding tools to the language model, defining agent and action nodes, and visualizing the graph with conditional routing.

Experiment with LangGraph tools by adjusting the input to avoid unnecessary searches. Rerun the test cell and review the debug logs and streaming outputs to confirm whether actions were taken.

Explore practice opportunity solutions in LangGraph by simulating tool calls, tracking message history, and using conditional logic to decide when to call search or calculator tools.

Define and add a custom LangGraph tool that returns the current date in ISO format, then build a graph and run it to see the date.

Define a new addition and subtraction tool to bind to our AI agent, add it to the list of tools, and test the agent with the new tool.

Define a new custom tool in LangGraph that performs addition and subtraction, and test it with queries like 123.5 plus 456.7 and 100 minus 33.5.

Expand your LangGraph workflow by integrating the Amadeus flight search API with multiple tools to fetch current date and flight offers for tailored origin, destination, dates, and passengers.

Build a hotel search tool with the Amadeus API, limited to New York City and Paris, configure check-in and check-out dates, update the tools list, and test the integration.

Learn how to add a hotel search tool using Amadeus, fetch hotels by city, pass IDs to search, and extract hotel name, price, and currency for New York City.

Combine flight search, tablet search, and date tools to build an ai booking agent that fetches news, security advisories, and the cheapest flights from a user’s origin to destination.

Test the ai agent booking tool via a Lang Graph workflow to obtain the latest news and search flights, noting the free tool has limited hotel search.

Test the AI agent booking tool within an agentic workflow. Explore tools to find flights from Toronto to Paris and restaurant recommendations for July 2025, then build a Gradio interface.

integrate the ai booking agent with gradio in langgraph to create a streaming travel assistant interface. explore flight searches and memory-enabled chat workflows.

Summarize the project of building generic AI workflows in a graph with AI agents, including nodes, edges, conditional edges, retrieval, memory, and tools, plus Gradio interfaces and integrated tools.

Build a data science AI team with CrewAI to automate a regression workflow - data cleaning, feature engineering, model training, and evaluation - to predict future units sold.

Explore the manual build, training, and evaluation of regression models, using pandas for data prep, EDA, preprocessing, and feature engineering, then compare linear and random forest regressors.

Predict weekly supplement units sold using regression on features like date, price, discount, and location, by loading data, imputing missing values, encoding categoricals, and evaluating with r-squared, MAE, and MSE.

Explore the fundamentals of machine learning regression, predicting continuous values like sales or salaries. Learn simple and multiple linear regression, the least squares method, and practical house-price examples.

Plot the linear equations y equals two three times x and y equals 15 minus ten x, map to linear regression with m and b, and estimate intercept and slope.

Practice regression basics by mapping linear equations to y = b + m x, identify slope and y-intercept, and compare lines with positive and negative slopes.

Import pandas, numpy, matplotlib, seaborn, and scikit-learn to inspect a pandas data frame; apply train_test_split, imputer, linear regression, and a random forest regressor, with MAE, MSE, and R2 metrics.

Inspect a 4300-row, eight-column pandas data frame; use df.head to view the first few rows, then df.info and df.describe to summarize data types, statistics, and missing values.

Learn to inspect a pandas dataframe by displaying the last ten rows, count unique products, and list the number of countries and platforms using code and external research strategies.

Learn to inspect data in Python with pandas by using tail to view the last rows, selecting a location column, and counting unique categories with nunique.

Learn practical strategies for managing missing data, including dropping rows and applying mean, median, and mode imputation with a simple imputer, ensuring price and discount columns are fully imputed.

Load raw data with pandas read_csv into a data frame, apply mean imputation, drop missing values with dropna, re-read the data, and perform a final row count sanity check.

Apply mean imputation with pandas and scikit-learn to fill missing price and discount columns, then compare with a dropna approach on the dataset.

Perform exploratory data analysis and visualization with matplotlib, seaborn, and plotly to explore distributions, histograms, scatter plots, heatmaps, interactive charts, and correlations in units sold.

Develop code to identify the strongest positive correlations with units sold from a correlation heat map, and compare platforms by median sales using box plots while plotting location counts.

Visualize a panda's data frame to identify the strongest positive correlation between units sold and discount. Compute platform medians of units sold and plot location counts with Seaborn for comparison.

Learn data preprocessing for machine learning by performing one-hot encoding with drop-first to avoid multicollinearity, select features, and split data into training and testing sets.

Experiment with a train-test split in pandas by toggling shuffle to true, using a 30/70 split, displaying the dataframe, and verifying that xtrain and xtest rows sum to original data.

Apply data pre-processing by splitting the dataset with train_test_split, using test_size 0.3 for a 70/30 split with shuffle enabled, and verify the shapes.

Build a linear regression model with scikit-learn, train on x and y to learn coefficients, then evaluate predictions on the test set using mae, mse, rmse, and r-squared.

Explore scikit-learn regression by evaluating r squared to explain variance in units sold and interpreting rmse for typical prediction errors, with quick code to analyze the results.

Demonstrate the practice solution by reporting R squared and RMSE for a linear model. Apply simple if-else rules to classify fit, and consider nonlinear models for underfitting.

Use scikit-learn to build a random forest regressor and train on the data. Compare to linear regression and evaluate with MAE, RMSE, R-squared, and a feature importance plot.

Train an XGBoost model to predict units sold and compare it to linear regression and random forest, install XGBoost, and research its power in Kaggle competitions.

Install and train an XGBoost regressor, compare r-squared and RMSE against linear regression and random forest, and explore hyperparameter tuning for better performance.

Analyze feature importance with a random forest regression model to identify key drivers of revenue, notably price and discount, and visualize results with a seaborn bar plot.

Split data into 75% training and 25% testing, retrain linear regression, random forest, and XGBoost, compare performance, and examine the impact of different splits and shuffling.

Explore feature importance analysis by comparing linear regression, random forest, and XGBoost to predict unit sales, using pandas for data prep and scikit-learn for model evaluation.

Learn to build a crew AI system that automates data science tasks—reading, cleaning, and splitting data; training regression models (including random forest); evaluating results—using role-based agents, tools, and sequential collaboration.

Install and import crew AI, openai, pandas, and lang chain; configure api keys and set GPT-4.1 mini. Run the notebook executor to validate a simple add function.

Create a function that takes two strings and joins them with an underscore, then run it using the notebook executor to test the result in this practice opportunity.

Explore how to test the notebook executor tool by writing a simple Python function that concatenates strings with an underscore, then run code to verify the tool outputs lm_engineering.

Define three ai agents—planner, data analysis and pre-processing, and modeler and evaluator—to execute a notebook workflow that builds a random forest regression model with 80/20 train-test split and feature importances.

Modify the analyst preprocessor agent to ensure train test split data is shuffled, update the prompt accordingly, and understand the agent definition for seamless ai workflow automation.

adjust an AI agent by enabling shuffle in a train test split (80/20) using the analyst pre-processor agent to control data division and agent performance.

Define key tasks and assign agents for a CrewAI workflow, covering planning, data analysis, and modeling evaluation with notebook execution for Python code.

assemble the crew of agents to generate and execute python code for a data analysis workflow, including preprocessing, modeling, evaluation, and reporting regression metrics.

Modify tasks to include a decision tree regressor, update agents and tasks, assemble the crew, run it, and evaluate and compare its performance against the random forest regressor.

Assemble a crew to train both decision tree and random forest regression models, compare their metrics, and generate code outputs with a clear definition of done.

Leverage crew AI to assemble agents and define tasks for data pre-processing, model training, evaluation, and reporting, automating the ml workflow to boost productivity while preserving human intuition.

Leverage the model context protocol (MCP) to connect AI agents with tools via a unified interface, enabling an MCP server with Gradio for a tutor with quizzes, summaries, and flashcards.

Discover the model context protocol (MCP), an open standard that lets AI models securely talk to tools, APIs, and data via a universal USB-C-like interface.

Install key libraries and configure api keys with gradio, openai, and python dot env, then set up an mcp server using server-sent events for an ai tutor.

Build and run the MCP server to expose AI-powered tools for explaining concepts, summarizing text, generating flashcards, and interactive quizzes, with streaming responses powered by OpenAI.

Continue building the MCP server by defining a generate flashcards function, streaming json-line q&a cards, and configuring the quiz me function for topic-based, level-aware questions, then prepare to launch Gradio.

Launch the gradio mcp server and expose four functions—explain concept, summarize text, generate flashcards, and quiz me—as tools on the mcp server, while using the gradio interface for testing.

Add a new tool to the MCP server that explains a concept in a different language, define its function, integrate it with Gradio, and run it again.

Add a new tool to the MCP server that explains a concept in a chosen language and level, via a function explain concept in language, tested with Gradio.

Explore tools on the MCP server and fetch the manifest schema via an http get request. Learn to describe tool inputs, including explain concept, summarize text, flashcards, and quiz me.

Stop the grading server in MCP servers, improve the core function's docstring to clarify the concept and level of detail, restart the server, and fetch the updated schema.

Update the explain concept docstring, insert a new cell in the MCP server manifest schema, and fetch an enhanced schema with detailed explanations.

Create an ai agent using the OpenAI agents sdk with mcp tools, featuring agents, handoffs, and guardrails. It explains a concept, summarizes text, generates flashcards, and quizzes users.

Explore how the MCP framework standardizes tool interactions via usb-c analogy. Transform tools into MCP services with Gradio, fetch manifest and schema, and build autonomous OpenAI sdk driven agents.

Explore low-code, drag-and-drop genetic AI workflows with N810, connecting Gmail, Google Sheets, and OpenAI APIs to create memory-enabled AI agents with tools and calendar integration.

Build your first agentic ai workflow in n810 by configuring a chat trigger, a summarization agent, and an OpenAI brain to summarize user input in one sentence.

Export and import ai agent workflows as json, manage variables, and track logs with visualizations of inputs, outputs, and tokens in N810.

Build an agentic AI workflow to automate translation using a cloud API, with a chat trigger, translating English to French as baseline, and test in the bootcamp.

Build and test a translation agent workflow using Claude, configure the AI model, set a custom prompt to translate input to a target language such as French, and verify results.

Configure an ai agent with OpenAI, add memory, and enable Serp API search using N810 templates to build a memory-enabled conversational workflow.

Practice opportunity to test the agentic AI workflow's search capability by querying bitcoin price and history since 2010, plus top five restaurants in the world, while logging memory and context.

Demonstrate an AI agent's search capabilities by querying Cirp and search APIs to fetch bitcoin circulation, current and historical prices, and top restaurants, while inspecting logs and memory.

Integrate Google Sheets into an ai agent workflow with n8n by appending rows that store the original input and its summary in a two-column sheet.

Develop an agentic ai workflow to convert python code to c, log versions in a Google sheet named programming language translation, and test with a function and a for loop.

Build a Python to C conversion workflow by configuring a Google Sheets document with Python as input and C as output, connected to an agent and tested with a sample.

Learn to generate parsed structured output in n8n using the structured output parser, producing JSON with tag and summary, and map results to Google Sheets.

Automate scheduling with Google Calendar workflows by building an AI agent that reads calendar events and creates new ones using get many and create event tools.

Add a Gmail trigger to the workflow to parse email content and schedule calendar invites with the AI agent.

Explore ai agents in production today, including Google Gemini's deep research and Microsoft Copilot's analyst agent, which plan, search, summarize, generate web pages, and produce Python visualizations from datasets.

Celebrate completing the course with hands-on experience building real-world AI agents. Stay connected, share projects on LinkedIn, and join the growing community of AI engineers shaping the future.