In this episode of the AWS Developers Podcast, we dive into the different ways to deploy large language models (LLMs) on AWS. From self-managed deployments on EC2 to fully managed services like SageMaker and Bedrock, we break down the pros and cons of each approach. Whether you're optimizing for compliance, cost, or time-to-market, we explore the trade-offs between flexibility and simplicity. You'll hear practical insights into instance selection, infrastructure management, model sizing, and prototyping strategies. We also examine how services like SageMaker Jumpstart and serverless architectures like Bedrock can streamline your machine learning workflows. If you're building or scaling AI applications in the cloud, this episode will help you navigate your options and design a deployment strategy that fits your needs.

With Germaine Ong, Startup Solution Architect ; With Jarett Yeo, Startup Solution Architect

Blog: Deploying Deepseek R1 Distill on Amazon EC2

Blog: Deploying DeepSeek R1 Distill on Amazon Sagemaker Jumpstart

Ollama

Open Web UI

Doc: deploy your own model on Amazon Sagemaker

Doc: deploy your own model on Amazon Bedrock