Introduction to Gen AI with AWS
An end-to-end walkthrough of deploying, evaluating and fine-tuning foundation models on AWS SageMaker and Bedrock.
Technical Information
Project Type
Cloud MLModel DeploymentFine-tuning
Tech Stack
AWSAWS BedrockSageMakerEC2S3Fine-tuningEndpointsLLMSLMPythonJupyter
Topics / Modules
Instance SetupModel EvaluationEndpoint ConfigurationFine-tuningS3 ArtifactsEndpoint Cleanup
Links
Project Summary
A practical project (Udacity domain-expert models) demonstrating the full lifecycle of generative-AI models on AWS – from spinning up instances and deploying endpoints to evaluating, fine-tuning and tearing everything down cost-consciously.
Skills Demonstrated
SageMakerBedrockModel DeploymentModel EvaluationFine-tuningCloud Cost Management
Solution
Using SageMaker and Bedrock, foundation models are deployed to endpoints, evaluated on domain data, fine-tuned, and the resulting artefacts stored in S3 – with careful endpoint cleanup to control cost.
Approach
- Provision: Create and run the required compute instances on AWS.
- Deploy: Configure and deploy model evaluation endpoints.
- Evaluate: Run evaluations and capture model outputs.
- Fine-tune: Fine-tune the model and persist artefacts to S3.
- Clean up: Delete deployment endpoints to manage cost.