AWS Certified Machine Learning Associate MLA-C01 - Updated

Learn to create your free AWS account, explore free tier products including EC2 and SageMaker, and understand billing limits and root account setup.

Learn to set up multi-factor authentication on the AWS root account with an authenticator app, scan the QR code, enter codes, and review IAM security settings.

Create and log in to an IAM user with an account alias, avoid root login, assign administrator access, and enforce MFA for secure AWS management console access.

Install and verify AWS CLI v2 on macOS or Windows, then configure credentials with AWS configure, create access keys, set the region to us east 1, and safeguard your keys.

explore identity and access management in AWS by defining users, attaching policies, and assigning permissions to enable or restrict actions, including service and execution roles.

Learn how data engineering designs scalable systems to collect, store, transform, and analyze data for analytics, using AWS data ingestion tools such as Kinesis Firehose, Snowball, and Storage Gateway.

Explore AWS data engineering tools, including data storage with S3, data integration with Glue, data warehousing with Redshift, and data visualization with QuickSight, to enable scalable, cost-efficient data processing.

Discover how to use AWS S3 buckets and objects, and compare storage classes like S3 Standard, S3 Intelligent-Tiering, S3 One Zone-IA, S3 Glacier, S3 Glacier Deep Archive, and S3 Outpost for scalable, durable storage.

Create a globally unique S3 bucket in the AWS Management Console, configure US East 1, enable ACL and bucket ownership, upload an object, and publish it with public access.

Learn to install and verify the AWS CLI v2, generate IAM access keys, configure credentials with aws configure, and set up named profiles for multi-account use.

Learn to create an S3 bucket with the AWS CLI, upload a file, and apply a lifecycle rule to move objects to Standard IA after 30 days for cost optimization.

Explore how AWS S3 intelligent-tiering automatically moves infrequently accessed objects to archive storage based on usage, without performance impact. Contrast this with lifecycle management, which moves data after 90 days.

perform a quick cleanup by deleting the file copied to the s3 bucket using the aws s3 mb command, then delete the s3 bucket to finalize the activity.

Enables data replication in S3 to achieve a recovery point by mirroring objects from source bucket to destination bucket within 15 minutes, with versioning and a configured IAM role.

Learn to secure Amazon S3 with least-privilege access, encryption at rest and in transit, MFA for delete, versioning, and object lock, plus auditing with CloudTrail, Macie, and GuardDuty.

Explore Amazon Kinesis for data analysis by ingesting, processing, and analyzing streams with Kinesis Video Streams, Kinesis Data Streams, Kinesis Data Firehose, and Kinesis Data Analytics, and compare to SQS.

Create and validate an AWS Kinesis stream to ingest streaming data, configure a shard, send a record, verify it via the data viewer, then delete the stream.

Create a Kinesis data stream and an S3 bucket, configure a producer Lambda to read S3 files and push to Kinesis, then deploy consumer Lambdas to process streamed data.

Create an S3 bucket and a Kinesis Data Stream, configure the Kinesis Data Firehose delivery stream to push streaming data to S3, and verify the setup with demo data.

Create an S3 bucket, a Kinesis Firehose delivery stream, and a Kinesis Data Analytics legacy SQL application to perform real-time analytics, delivering results to S3 and testing the destination.

Explore AWS Glue, a serverless data integration service that discovers, cleans, transforms, and loads data from 70+ sources into a data lake, with a centralized catalog and ETL pipelines.

Discover how AWS Glue crawlers automatically discover metadata and schema from a CSV file in S3, create a Glue database and tables, and populate a data catalog for easy querying.

Learn how to transform data with AWS Glue DataBrew by creating a sample project, applying upper-case formatting to a CSV dataset, and exporting the transformed data to S3.

Perform an ETL operation in AWS Glue using S3 as data source, with a crawler creating a table in the Glue data catalog for later aggregation in an ETL job.

Discover Amazon Athena, a serverless, interactive query service that analyzes data stored in S3 with standard SQL, enabling ad hoc queries in parallel and pay-as-you-go pricing.

Configure Amazon Athena to query S3 data without indexing, create a data catalog, database, and table, upload case.csv to the S3 bucket, and run SQL queries to analyze the data.

Discover how AWS Batch dynamically provisions resources to run hundreds of thousands of batch jobs, with Fargate and GPU scheduling, serverless options, and integrated monitoring and logging.

Learn how AWS step functions provide serverless orchestration with a visual workflow to manage standard and express workflows, including exactly-once and at-least-once executions and branching and error handling.

Learn to build and test a step function workflow by creating a Python lambda, deploying code, wiring the lambda to a state machine, and executing a sample run.

Learn to build a serverless workflow with aws step functions that orchestrates lambda and services, manages state and retries, and uses dynamodb and sns for end-to-end automation.

Explore AWS Step Functions by building a five-lambda workflow that adds two numbers, checks if the sum is greater than ten, prints or increments, waits, and tests the state machine.

Explore how to author end-to-end ML pipelines on AWS Step Functions using Python, drag-and-drop, and JSON, integrating with SageMaker, Lambda, and S3 for training, model creation, and deployment.

Explore data pre-processing with AWS Glue DataBrew and AWS Step Functions to train an ML model on an air quality dataset that predicts NYC CO2 levels.

Explore data with statistics and visualization to identify patterns, outliers, missing values, and imputation needs, and apply pca and lda using Apache Spark, EMR, QuickSight, and Athena.

Perform hands-on exploratory data analysis on a cars data set using Pandas and Seaborn to inspect MSRP distribution, features, and missing values.

Identify and distinguish qualitative and quantitative data types, including nominal and ordinal, discrete and continuous. Explain how to classify features and dependent versus independent variables for data analysis and EDA.

Explore descriptive, inferential, predictive, and prescriptive analyses, along with exploratory data analysis and causal analysis, to extract patterns and insights from data using mean, median, and mode.

Master descriptive statistics to summarize data using mean, median, mode, and measures of variability. Learn frequency distribution, central tendency, and univariate to multivariate analysis with practical examples.

Identify outliers using box plots and IQR rules (Q1, Q3, 1.5 IQR), and examine distributions with histograms and bar charts to support exploratory data analysis.

Perform exploratory data analysis by acquiring data from csv, excel, and sqlite, then merge into a single data frame using pandas and sqlalchemy.

Clean data by removing duplicates, save the cleaned data to csv, and reorder columns for clarity. Perform outlier analysis with box plots and five-number summaries to drop anomalies.

Learn to handle missing values by dropping or imputing data, using median for numeric columns and mode for categorical ones, and visualize missingness with pandas and the missing number library.

Impute missing values and fix categorical typos by applying custom functions to product type, rating, and payment mode, then prepare the data for transformation.

Explore data transformation for exploratory data analysis by converting date columns to date time, cleaning price data to numeric INR, and handling currencies, missing values, and outliers with IQR.

Encode categorical data using label encoding for ordinal data and one-hot encoding for nominal data with sklearn and pandas get dummies. Use rating and shipping mode as examples.

Scale numerical data using standard scaler and min-max scaler to normalize features, achieve zero mean and unit variance, and apply these methods on columns like price INR.

Explore exploratory data analysis visualizations for HR attrition data, using histograms, count plots, bar plots, swarm and box plots, and scatter/joint plots to reveal attrition patterns.

Learn how imbalanced datasets affect classification, and balance data using undersampling, oversampling, SMOTE, and near miss with imbalanced learn library in fraud and credit card scenarios.

Explore principal component analysis to reduce dimensionality while preserving variance. Apply standardization, compute covariance, and project data using eigenvectors into a lower dimension.

Learn linear discriminant analysis, a supervised dimensionality reduction method that uses both features and target to maximize class separation, with two components and a scikit-learn implementation.

Explore Amazon QuickSight, a cloud-based, fully managed visualization tool that connects data from AWS and other sources to create fast, interactive dashboards, with ML insights and anomaly detection.

Understand that machine learning learns from data by pairing inputs with outputs to train a model, contrasting it with traditional software, and applying features to tasks like spam detection.

Explore the three main types of machine learning: supervised, unsupervised, and reinforcement learning, and how labeled data, features, and patterns guide model selection, with a preview of linear regression.

Apply linear regression to predict sales from TV, radio, and newspaper spend, train on a split dataset, and evaluate with mean absolute error, mean squared error, RMSE, and R2.

Apply L1, L2, and elastic net regularization to linear regression, showing how penalties affect coefficients and model performance. Assess assumptions, multicollinearity with VIF, and residual diagnostics.

Explore how logistic regression governs binary outcomes by applying a logistic function to linear regression outputs, providing probabilities and enabling binary classification via thresholding.

Explore gradient descent, a foundational optimization method in machine learning and deep learning, showing how to update parameters using loss gradients and learning rate to train models like logistic regression.

Implement logistic regression on a bank dataset to predict term deposit subscriptions. Use one-hot encoding and SMOTE, then apply recursive feature elimination to reach 0.88 accuracy.

Evaluate classification models using accuracy, confusion matrix, precision, recall, and f1 score; then assess discrimination with roc auc and probability-based thresholds.

Explore how the decision tree algorithm splits data using questions, forming root, branches, and leaves in supervised learning for classification and regression, guided by impurity-based loss like Gini or entropy.

Explore how decision tree loss functions guide splits, comparing entropy and information gain with gini impurity, and reviewing algorithms such as ID3, C4.5, CART, CHAID, and Mars.

Implement a decision tree classifier on the social network ads dataset, from data loading and scaling to training, predicting, and evaluating with accuracy and F1, and visualize the resulting tree.

Understand overfitting and underfitting and how data noise and model complexity affect them. Learn how k-fold cross validation prevents overfitting and improves generalization.

Learn hyperparameter optimization techniques for decision trees, including manual, grid search, and randomized search, with cross-validation to reduce overfitting and improve model performance.

Investigate the k nearest neighbor algorithm (kNN), a lazy, instance-based supervised method used for classification and regression, leveraging the nearest data points for predictions.

Learn how support vector machines classify data and perform regression by finding the maximum margin hyperplane, using linear and non-linear SVM with kernel tricks like polynomial, RBF, and sigmoid.

Explore ensemble learning and the wisdom of crowd to improve predictions by aggregating classifiers. See hard and soft voting with models like logistic regression, SVM, and random forest.

Get practical insight into bagging classifiers, built from bootstrap sampling and aggregation, training many models in parallel and aggregating votes to improve accuracy, including random forests and feature importance.

Explore boosting in ensemble learning, including AdaBoost with sequential models and weights on misclassified instances, and gradient boosting using residuals and summed predictions to reduce underfit and overfit.

Explore implementing gradient boosting with XGBoost, including the D matrix and training parameters for binary classification. Compare its performance and monitoring via ROC AUC to scikit-learn’s gradient boosting.

Explore unsupervised learning with clustering, focusing on k-means, exclusive clustering, internal cohesion, external separation, and metrics like inertia and silhouette score.

Explore hierarchical clustering, building a dendrogram from bottom up via agglomerative and divisive methods. Compare linkage metrics—single, average, complete, mean, centroid—using Euclidean distance to reveal cluster structure.

Learn dbscan, a density-based clustering that finds arbitrary-shaped clusters, is robust to outliers, requires no predefined cluster count, uses epsilon and min points, with a practical sklearn example.

Explore time series analysis, including trend, seasonality, cyclic patterns, and irregularity, and apply stationary testing with the augmented Dickey-Fuller test or the Kwiatkowski-Phillips-Schmidt-Shin test, plus differencing and ARMA forecasting.

Learn hands-on time series analysis with arima: convert non-stationary data with differencing, validate stationarity with ad fuller test to tune p, d, q using auto.arima, and generate forecasts.

Explore how Amazon Personalize builds and deploys machine learning powered recommendations and intelligent user segmentation at scale, delivering personalized movie and product experiences for customers.

Explore deep learning, a subset of machine learning, where artificial neural networks learn patterns from data through input, hidden, and output layers with activation functions and real-world applications.

Explore the basics of TensorFlow and tf.keras, build a first neural network with a sequential model, dense layers, activations, and the fashion mnist dataset.

Explore how a neural network with input, hidden, and output layers handles concentric circle data with sigmoid activation. Learn about forward and backward passes and gradient descent.

Explore activation functions in deep learning, from sigmoid and tanh to ReLU variants like leaky ReLU, ELU, SELU, swish, in TensorFlow, and use softmax for multiclass classification.

explore the common deep learning network architecture for binary and multi-class classification and regression, detailing input features, hidden layers, output neurons, activations, and loss functions.

Understand how epochs, batch size, and iterations govern training a neural network, including forward and backward passes, with practical examples showing how batch size and epochs shape learning and iterations.

Discover how to tune deep learning hyperparameters, including learning rate, batch size, and epochs, along with model choices like hidden layers, number of units, activation functions, and early stopping.

Explore vanishing and exploding gradients in backpropagation and learn remedies like Xavier initialization, ReLU activations, batch normalization, and gradient clipping to stabilize deep learning training.

Discover how computer vision uses convolutional neural networks to enable machines to interpret images. Apply kernels, padding, strides, and pooling to extract features and build a cat-dog classifier.

Implement a convolutional neural network on the cat dog dataset to build an image classifier, covering data loading, augmentation, the convolutional neural network architecture, and transfer learning techniques.

Explore how transfer learning leverages pre-trained models like ResNet with ImageNet weights to classify new image tasks, using feature extraction and fine tuning to adapt to new data.

Explore how feed forward neural networks learn input-output patterns, their limitations for text data and NLP, and the need for recurrent neural networks to remember prior context.

Explore recurrent neural networks and architectures—from 1 to 1 to many to many, including encoder-decoder, and learn about vanishing and exploding gradients along with remedies like LSTM.

Explore how long short term memory networks solve long term dependencies in sequences using forget, input, and output gates, and introduce attention-based encoder-decoder architectures for many to many tasks.

Explore the attention mechanism for sequence-to-sequence learning, replacing fixed-length encoder bottlenecks with attention over encoder states to improve neural machine translation.

Discover how transfer learning accelerates deep learning across vision and natural language tasks, using pre-trained bodies and heads, language modeling, domain adaptation, and fine-tuning.

Explore transformer architecture, including encoder and decoder components, positional encoding, multi-head attention, residual connections, and feedforward layers, to enable parallelized language modeling and accurate natural language processing predictions.

Learn how MLOps standardizes and streamlines the machine learning lifecycle from data preparation to deployment and monitoring, enabling cross-functional teams to deliver production models with governance and value.

Explore ml ops fundamentals across dev, staging, and prod environments, with code in git, a model registry, and an AWS ci/cd pipeline using SageMaker.

Explore mlops fundamentals with a deep dive into ci/cd pipelines, data preparation, feature store, model training and deployment, and monitoring for data drift and retraining.

DevOps alone cannot optimize ML projects; ML demands iterative experimentation, data governance, specialized infrastructure, and MLOps practices to effectively deploy learning models.

Understand Amazon Web Services and its cloud computing, with pay-as-you-go access to compute, storage, and analytics. Learn IaaS, PaaS, SaaS, deployment models, regions, and availability zones supporting scalable MLOps.

Explore the AWS technical stack for ml and MLOps, from data preparation and feature engineering to training, deployment, governance, and model registry, powered by SageMaker and pay-as-you-go services.

Explore Amazon SageMaker as a fully managed ML service for building, training, and deploying models with SageMaker Studio, MLOps tools, and SageMaker Canvas, while mastering notebook workflows and model lineage.

Explore how Amazon SageMaker separates notebook from infrastructure to manage data loading, preprocessing, experiments, and deployment, with scalable training, tracking, and real-time APIs via S3.

Deploy the trained cat-dog image classifier as a serverless aws lambda service using tf lite, with preprocessing, docker packaging, and api gateway integration for scalable web inference.

Discover how docker containers isolate applications, package dependencies with a dockerfile, and run, test, and deploy images across environments using docker pull, run, and port mapping.

Build a docker image from a TensorFlow Lite project, run it locally, and deploy to AWS Lambda via ECR, then test with HTTP requests and expose model via API gateway.

Explore how CloudWatch monitors resources and applications, creates alarms, visualizes metrics on dashboards, and automates actions through logs, log groups, and EventBridge rules.

Explore SageMaker Jumpstart to access pre-trained open source models, enable quick transfer learning, and deploy endpoints for machine translation and image classification with ready-to-use notebooks and documentation.

Explore AWS Polly, a text-to-speech service using deep learning to generate lifelike speech, with neural engine and standard options, and experiment with voices like Sally.

Learn end-to-end deployment of a SageMaker model: from notebook creation and data on S3, to training, deploying an endpoint, and invoking it via Lambda and API Gateway.

Explore amazon transcribe to convert speech to text, enabling real-time transcription, vocabulary filtering, custom language models, and post-call analytics for media search and monetization.

Learn how AWS Lex enables voice and text chatbots with templates and intents like booking flight and get flight status, compares to Dialogflow, and integrates transcribe, SageMaker, and Polly.

Explore Amazon augmented AI, which adds a human in the loop to review predictions from Rekognition and Textract via defined workflows and human loops, using confidence thresholds or random sampling.

Discover Amazon CodeGuru, an intelligent ML-driven service that improves code quality and performance for Java and Python apps, using CodeGuru profiler to visualize hotspots and CodeGuru reviewer to validate code.

Explore Amazon Comprehend and Amazon Comprehend Medical to perform natural language processing on text data, extracting entities, topics, key phrases, sentiment, and syntax with real-time analysis and custom models.

Learn machine learning with AWS DeepComposer, composing original music with pre-trained models via keyboard or virtual console across rock, pop, jazz, classical, or custom genres, and sharing tracks on SoundCloud.

Explore AWS DeepLens, a deep learning enabled video camera for computer vision with pre-built models and tutorials; register device and deploy models via the AWS management console for Lambda inference.

Learn hands-on reinforcement learning with AWS DeepRacer, a 1/18 scale autonomous race car, using a 3D simulator, and push models to the car for experimentation.

Explore how Amazon DevOps Guru uses ML-powered insights to automatically detect operational issues, improve application availability, and reduce downtime through actionable recommendations.

Explore Amazon Forecast for time series analysis and forecasting models powered by machine learning, using historical data and automated machine learning, the same technology used by Amazon.com.

Explore Amazon Fraud Detector, a fully managed AWS service that detects online fraud such as payment fraud and fake accounts, with out of box integration, data validation, and model deployment.

Explore Amazon HealthLake, a HIPAA eligible AWS service that stores, transforms, queries, and analyzes unstructured health data at petabyte scale using machine learning.

Explore Amazon Kendra, an AWS enterprise search service powered by machine learning that connects data sources with Kendra connectors. Enhance enterprise search with natural language queries.

Amazon Lookout for Equipment, Metrics, and Vision analyzes sensor data to detect anomalies in industrial assets. It supports up to 300 sensors, provides historical logs, and enables automated quality inspection.

Discover Amazon Monitron, an end-to-end system that detects abnormal behavior in rotating equipment and reduces unplanned downtime, with easy sensor and gateway installation via the Monitron app.

Explore AWS panorama, a machine learning appliance and SDK that adds computer vision to existing on-premise cameras, enabling edge predictions with low latency, data privacy, limited bandwidth, and automated tasks.

Explore Amazon Rekognition, a deep learning visual analysis service that searches, verifies, and organizes millions of images and videos with object detection, facial analysis, and celebrity recognition.

Explore Amazon Translate, a service for real-time and batch translation with API integration, and tailor translations using custom terminology and parallel data with translation memory.

extracts text and data from handwriting and scanned documents using machine learning, going beyond OCR to analyze forms, tables, and ID documents with the AnalyzeID API.