Copilot & AI Agents for Data Science Bootcamp [2026]

Learn data wrangling and feature engineering with Microsoft Copilot and pandas, using plain English prompts to clean, sort, merge, and normalize data for downstream machine learning.

Explore data wrangling and feature engineering with real-world datasets, using Copilot to clean, format, and organize data, handle missing values, and create powerful features for machine learning models.

Master prompt engineering for generative ai by crafting context, instructions, input data, and output indicators. Learn top tips: avoid hallucinations, show reasoning, set constraints, iterate, and use few-shot prompts.

Explore zero-shot prompting, few-shot prompting, and chain-of-thought prompting to guide Copilot and AI models, using examples, step-by-step reasoning, and sentiment and math tasks.

Explore the pandas library and its integration with Microsoft Copilot to read Excel data, handle large data, and enable reproducible analysis with visualization libraries matplotlib, seaborn, and Plotly Express.

Learn to import Excel files into pandas data frames with Copilot, perform data wrangling, merge demographics and financial details by client ID, and prepare data for analysis.

Locate and handle missing data in Pandas dataframes by using isnull, count with sum, and fill with mean or mode, demonstrated in Jupyter and Copilot prompts.

Merge two pandas data frames on client ID to create a unified clients combined data set, handling missing values and using Copilot for seamless analysis.

Learn to perform data analysis, filter and sort a pandas data frame, generate a statistical summary with describe, and identify high net worth clients by portfolio size above 150,000.

Learn data visualization with Matplotlib, Seaborn, and Plotly to create static and interactive charts, including scatter plots, histograms, bar charts grouped by advisor, and correlation heatmaps.

Master feature engineering techniques for data science, including imputation, outlier handling, binning, and transformations. Learn one hot encoding and scaling methods—normalization and standardization—for preparing data for machine learning.

Explore data wrangling and feature engineering on HR data to predict attrition using Microsoft Copilot, loading the dataset, imputing missing values with mean/median, and applying one-hot encoding, standardization, and normalization.

Learn to perform one hot encoding of the education field using pandas get_dummies, then standardize age and normalize monthly income with scikit-learn, with before-and-after checks.

Filter a pandas data frame to select loyal employees with at least 30 years in R&D. Create stayed and left dataframes, summarize attrition factors, visualize tenure with box plot.

Build a custom AI agent in Copilot to perform data wrangling and analysis on real-world order data, using feature engineering and visualizations, and explore GPT-5 features.

Build a data wrangling ai agent in copilot that loads csvs, performs eda, cleans duplicates and missing values, and saves a verified clean dataset after user confirmation.

Use Microsoft Copilot to wrangle Walmart sales data, impute missing values, summarize key metrics, and visualize total sales by department, while applying standardization, normalization, and one-hot encoding for store location.

Load Walmart sales data into a pandas data frame, examine shape and statistics, then impute missing values by mean for unit price and mode for customer segment, and verify completion.

Practice data wrangling and feature engineering with Copilot to generate data visualizations and code. Explore standardization, normalization, one-hot encoding, and insights like total sales by department and location.

Build a data wrangling AI agent in Microsoft Copilot to automate data cleaning and preparation with pandas, applying one-hot encoding, normalization, and standardization along with zero-shot, chain-of-thought, and few-shot prompting.

Explore data visualization and storytelling with Microsoft Copilot and GPT five, using Python-generated visuals from matplotlib, seaborn, and plotly to analyze happiness and sales datasets.

Explore various data visualization types, including relationships (scatter plots and bubble charts), comparisons (bars and lines), distributions (histograms and box plots), and compositions (pie, stacked bars, and stacked area charts).

Leverage Microsoft Copilot to visualize World Happiness Report data in Excel, exploring happiness score, GDP per capita, social support, freedom, generosity, and corruption for storytelling.

Leverage Copilot to generate Python visualizations with Matplotlib and Seaborn: a scatter plot of GDP per capita vs happiness score with a best-fit line, and a top-ten annotated bar chart.

Visualize a scatter chart of social support and happiness score, with a regression line. Create a horizontal bar chart showing the ten lowest happiness score countries.

Shows solving a practice opportunity by visualizing social support versus happiness with a regression line in a scatter plot, and a horizontal bar chart of the top ten lowest-scoring countries.

Create correlation heatmaps and Seaborn pairplots from the happiness report data, using pandas and seaborn, and generate ten GPT-5 visualizations with matplotlib and Plotly.

Leverage the analyst AI agent in Microsoft Copilot to generate visuals from the happiness data, including histograms, top ten happiest countries, GDP per capita comparisons, and a correlation heat map.

Visualize Walmart sales data with Copilot and the analyst AI agent to create bar charts, box plots, scatter plots, and heat maps of weekly revenue, visits, stock levels, and weather.

Visualize Walmart weekly revenues across stores with line charts and heatmaps using Copilot, revealing store-level trends, top revenue leaders, and the strong link between customer visits and sales.

Explore how weather conditions and temperature influence Walmart store revenue through regression analysis and visualizations, using Copilot-assisted data science workflows.

Practice using the analyst AI agent in Microsoft Copilot to generate twenty data visualizations with Seaborn, Matplotlib, and Plotly Express, exploring varied insights from the data.

Explore 20 distinct data visualizations for Walmart sales data using the Copilot AI Analyst Agent, leveraging matplotlib, seaborn, and Plotly express to reveal correlations, distributions, and city revenue trends.

analyze and visualize cancer.csv using seaborn, matplotlib, and Plotly to compare benign and malignant samples, generating multiple visuals and exploring correlations to tell a story from the data.

Unlock cancer data storytelling through Copilot guided analysis, building pandas data frames and creating seaborn visuals—histograms, KDEs, box plots, and pair plots by target class.

Plot cancer data including bar charts of feature correlations to the target, a benign-vs-malignant pie chart, count plots, scatter and heat map plots using pandas, matplotlib, seaborn, and plotly express.

See how the Copilot analyst AI agent analyzes the cancer dataset to generate 20 visualizations, including histograms, box plots, violin plots, a pair plot, and a color-coded correlation heatmap.

Master data visualization with matplotlib, seaborn, and Plotly Express, automate visuals with Copilot Analyst Agent, and build regression line plots to study feature correlations for decisions.

Learn to develop and validate models using Microsoft Copilot, compare classical classifiers and neural networks, evaluate with confusion matrix and metrics, and visualize performance with roc curve and auc.

Use copilot to build a marketing data pipeline with eda and visualization, train binary classifiers to predict campaign responses, score leads with confidence, and recommend actions to marketing teams.

Compare logistic regression, decision trees, SVM, and random forest classifiers, focusing on probabilities, decision boundaries, interpretability, performance, and practical use cases like credit scoring and fraud detection.

Explore how a confusion matrix visualizes classifier performance, showing true positives, true negatives, false positives, and false negatives, and learn how accuracy, precision, and recall are calculated.

Explore evaluating a fraud-detection model with a confusion matrix on 1000 credit card transactions, computing accuracy, precision, and recall, and decide which metric the bank should prioritize and why.

Learn how to compute accuracy, precision, and recall from a confusion matrix (92%, 58.3%, 70%) and why banks prioritize precision to minimize false positives while safeguarding fraud detection.

Build an AI agent in Copilot to manage the full data science lifecycle, from data ingestion and EDA with visuals to one-hot encoding and predicting campaign responses.

Build an ai agent in Microsoft Copilot Studio by assembling prompts in stages, configure a marketing data scientist, and run data cleaning, feature engineering, one-hot encoding, modeling, and evaluation.

This practical project guides you through building a marketing data scientist agent in CoPilot, running prompts to summarize data, generate visuals, and explore features for predicting campaign conversion.

Train and compare logistic regression and random forest classifiers, evaluate with roc auc and confusion matrices, review top contributing features, then generate predictions for unseen leads and export to Excel.

Train and assess multiple classifier models in Copilot, from logistic regression to neural networks, using metrics like accuracy, precision, recall, F1, ROC AUC, and confusion matrices; compare and interpret results.

Practice opportunity solution part a trains and compares multiple ml classifier models in Copilot, from logistic regression to neural networks, with eda and auc evaluation.

Explore how a pre-built analyst ai agent trains ten classifier models, preprocesses data, and uses roc curves and auc to compare performance, revealing a leaderboard and tuning options.

Recap how data, algorithms, and resources shape machine learning classifier development, from logistic regression to Naive Bayes, evaluate metrics like ROC/AUC, and use Copilot to build, evaluate, and interpret pipelines.

Explore anomaly detection concepts and applications using Microsoft Copilot, learn the z-score method and isolation forest, and build sales anomaly detection agents in practical demos.

Learn the basics of anomaly detection, identify data points that deviate from the norm, and apply z-score and isolation forest techniques for outlier detection.

Explore the z-score method for anomaly detection by measuring how far a data point lies from the mean in standard deviation units, and apply a threshold to flag outliers.

Build an anomaly detector AI agent in Microsoft Copilot using the z score method on sales data, identify anomalies with a plus or minus two threshold, and interpret business implications.

Build and evaluate a sales anomaly detector AI agent in Copilot. Use pandas and matplotlib to plot data, compute mean, standard deviation, and z-scores, then identify and visualize anomalies.

Explore the isolation forest algorithm for unsupervised anomaly detection, using a forest of decision trees to isolate outliers and compute anomaly scores.

Build an AI agent in Microsoft Copilot called Isolation Forest Detector to detect monthly sales anomalies with isolation forest algorithm; visualize results, interpret business impact, and compare with Z-score method.

Build an ai agent in Microsoft Copilot to detect anomalies with the isolation forest algorithm, visualize results with plots, and compare against the z-score method.

Explain anomaly detection concepts and real-world applications, including z score outlier detection in sales data, and compare Copilot AI agents with isolation forest for performance and business impact.

learn simple linear regression to predict y from x, using the straight-line model y = m x + b, with slope m and intercept b, to forecast revenue from temperature.

Explore least sum of squares to estimate the linear regression parameters m and b, using residuals and y-hat, with a temperature revenue example, and learn train-test data splitting for generalization.

Explore scikit learn, a free Python machine learning library, covering regression, classification, clustering, and preprocessing, including train/test splits, scaling, and linear regression.

Explore the XGBoost algorithm through gradient boosting, building sequential models to predict residuals, applying learning rate and regularization to improve ensemble predictions.

Explore XGBoost, the extreme gradient boosting algorithm, and learn how ensemble of weak models yields strong predictions for regression and classification, with hyperparameter tuning and practical trade-offs.

Explore boosting, an ensemble method that trains weak models in sequence, learns from residuals, reweights data to improve predictions, with a practical XGBoost example and discussion of extrapolation limits.

Explore ensemble learning, including bagging and boosting, and see how multiple decision trees combine to improve accuracy and generalization in XGBoost, with leaves, decision nodes, and CART concepts.

Explore the bias-variance trade-off using training and testing data, showing underfitting and overfitting with a simple linear model versus a high-order model, and balance bias and variance to generalize well.

Explore L2 regularization, ridge regression, to balance bias and variance, reduce overfitting, and improve generalization across training and testing data, with alpha as the tuning parameter in models like XGBoost.

Explore L1 regularization (lasso), where the alpha-weighted absolute slope penalty reduces overfitting and enables feature selection, in contrast to L2 ridge with squared penalties.

Define data strategy foundations with data acquisition, exploration, cleaning, and feature engineering. Align governance, privacy, compliance, storage, and data pipelines to drive ai model performance and kpis.

Apply a practical data strategy to predictive maintenance by integrating sensor data and maintenance logs. Learn data discovery, collection, quality, governance, privacy, storage, labeling, integration, and KPIs to cut downtime.

Define data requirements to train, validate, and deploy AI models, covering data type, structure, sources, volume, velocity, and labeling, with structured vs unstructured and labeled vs unlabeled data.

Differentiate supervised learning with labeled data, unsupervised learning with pattern discovery, and reinforcement learning through agent–environment interactions and rewards and penalties, illustrated by image labels and clustering.

Compare internal vs external data sources and use publicly available datasets to expand your data, with examples from Hugging Face, Bloomberg API, and Kaggle.

Assess data quality by distinguishing good data from bad data across relevance, representation, accuracy, completeness, balance, consistency, timeliness, privacy, and signal-to-noise; illustrate with spam detection.

Discover how data labeling tags images, text, audio, and video for supervised learning using guidelines and a mix of automated and human labeling workflows on Amazon SageMaker Ground Truth.

Compare database, data warehouse, and data lake storage: schema-based transactional databases, analytics-focused data warehouses, and flexible unstructured data lakes for BI, analytics, and machine learning outputs.

Explore how data pipelines ingest, validate, transform, store, orchestrate, and consume data for analytics while applying governance and security across the AI lifecycle.

Compare Microsoft Copilot, Copilot Pro, and Microsoft 365 Copilot, detailing free versus paid plans, real-time results, model access, grounding, and Copilot Studio for ai agents.

Explore Microsoft Copilot, an AI assistant across PowerPoint, Teams, Word, Outlook, and Excel to generate slides, summarize meetings, write documents, and create charts with AI agents.

Explore Copilot's general use cases: analyze financial statements, compute liquidity ratios, extract data from PDFs, and automate presentations and reports in PowerPoint, Excel, Word, and Outlook.

Learn to perform data wrangling in Excel with Python via Copilot’s advanced data analysis to clean data, generate insights through dashboards, and build machine learning models within a spreadsheet.

Celebrate completing the bootcamp and apply hands-on with Microsoft Copilot and AI agents for data science, while connecting with peers on LinkedIn and joining a growing AI engineering community.