Amazon SageMaker

Take a tour of the AWS SageMaker console to explore its services and end-to-end machine learning lifecycle, and learn how to set up a SageMaker domain and Studio.

Explore how Amazon SageMaker domains and users organize studio environments, detailing domains, user profiles, shared spaces, and role-based permissions for data scientists, ML ops, and compute tasks.

Create a SageMaker studio user profile linked to a domain, assign roles, enable jumpstart and projects, then explore studio's home with Jupyter and VS Code.

Launch a preloaded SageMaker Studio JupyterLab notebook environment with libraries like PyTorch, TensorFlow, Keras, NumPy, and pandas, enabling unified analytics and ML workflows with Code Whisperer and PySpark options.

Explore Amazon SageMaker Canvas, a no-code interface that builds accurate ML models with ready-to-use and custom options, data prep via Data Wrangler, and natural language or point-and-click interactions for ML.

Explore SageMaker Data Wrangler to quickly prepare tabular and image data with 300+ PySpark transformations, built-in data quality insights, visualizations, and code-free pipelines for ML.

Explore Data Wrangler in Canvas within SageMaker, create and import datasets, build data flows, handle missing data and outliers, visualize analyses, and export cleaned data for model building.

Run a machine learning model in Amazon SageMaker Canvas using the clean penguins data from Data Wrangler, with validation, export, preview, build, predict, and deploy.

Explore ready-to-use text models in SageMaker canvas for sentiment analysis, language detection, entity extraction, and personal information detection with single or batch predictions.

Explore ready-to-use models for images and documents, leveraging object and text detection, bounding boxes, and Textract-powered expense analysis and document queries.

Deploy canvas models by creating a deployment, selecting the model version, and choosing the instance type and count. Activate the endpoint for real-time inferences and monitor costs.

Explore Amazon SageMaker Jumpstart, a gateway to ready-to-use pre-trained open source foundation models via notebooks and solution templates, with Studio and Python SDK options for quick, scalable ML deployment.

Explore SageMaker Jumpstart in studio, configure execution roles and permissions to deploy large models, set up endpoints, and run example notebooks with Mistral 7B instruct.

Learn to use the SageMaker Python SDK to prepare data with pandas and seaborn, upload to S3, and train and deploy a k-nearest neighbors model in SageMaker Studio.

Explore Amazon Bedrock, a fully managed gen AI service unifying foundation models from Cohere, Anthropic, Meta, Stability AI, and Titan via Boto3 in Python, including text and image generation.

Learn to connect Python with AWS via boto3 by installing boto3, configuring IAM access keys, and managing credentials in the dot AWS folder.

Explore text generation parameters in Amazon Bedrock models, including max token length, temperature, top P, and stop sequences, and learn how model variability affects JSON configuration and coding.

Learn to connect to the Amazon Titan text model with boto3, invoke bedrock runtime, and tune generation using temperature, top_p, and max_tokens.

Learn to call the llama two model via Amazon Bedrock and adapt to model-specific parameters. Observe prompt vs input text, camel vs snake case, and varying request shapes by model.

Explore image generation with Amazon Bedrock using Titan and Stability AI XL models, learning key parameters like text prompts, CFG scale, seed, and image sizing.

Learn how Stability AI diffusion models on AWS Bedrock generate images from text prompts using positive and negative prompts with weights, adjustable steps, samplers, CFG scale, seed, and size rules.

Explore the Titan image generator in Amazon Bedrock, generating 512x512 images from text prompts, adjusting quality to premium, and using image variation to stylize an apple on a table.

Learn Rag, retrieval augmented generation, by embedding text into vectors and using cosine similarity to retrieve relevant documents. Augment prompts with that content for more accurate responses.

Learn how to build retrieval augmented generation workflows by embedding documents, storing vector embeddings, and performing cosine similarity to retrieve context for answering questions with Amazon Bedrock.