Hello Reader,

In the previous newsletter edition, we took a look at top Gen AI tools, and how can you benefit from the trend. The matter of the fast is, even in general SA interviews, you have to expect Gen AI questions. This is similar to how you expect fundamental containerization and DevOps questions. Gen AI is becoming quite popular, and this is no exception. In today's edition, let's go over AWS Gen AI landsacape, that I am following:

The image illustrates the AWS Generative AI (Gen AI) landscape, categorizing services and tools across several layers.

Embedding and RAG (Retrieval-Augmented Generation)

This layer focuses on embedding data and enabling retrieval-augmented generation for AI applications.

• Bedrock (Knowledge Base): AWS Bedrock provides foundational models for generative AI tasks, allowing users to integrate these models into their applications without managing infrastructure.
• OpenSearch Vector Database: This service allows storing and querying vector embeddings, enabling semantic search and similarity matching in generative AI workflows.
• Neptune for graphRAG Vector: AWS Neptune is a graph database service that supports graph-based retrieval-augmented generation by connecting data relationships and embeddings.
• Embedding Models (Amazon Titan, Cohere Embed): These pre-trained models generate vector embeddings for text or other data types, facilitating semantic understanding and retrieval. Note that, these are NOT same as non-embedding tasks. You embed using these models, and can use a totally different model for actual prompt.

Agentic AI

This layer supports building intelligent agents capable of executing tasks autonomously.

• Bedrock (Using Action Groups): Extends Bedrock’s capabilities by enabling agents to take actions based on predefined workflows or decision-making logic.
• Lambda (Invoke from Agents): AWS Lambda allows serverless execution of functions triggered by agent actions, enabling scalable task automation.
• SageMaker AI: A machine learning platform for training, deploying, and managing custom AI models that can be integrated into agentic workflows.

Infrastructure

This layer provides the foundational hardware and services to train and host AI models.

• EC2 (Trainium, Inferentia, GPUs): AWS EC2 instances equipped with specialized hardware like Trainium chips, Inferentia processors, and GPUs are optimized for training and inference of large-scale AI models.
• SageMaker AI: Besides its role in agentic AI, SageMaker also serves as a robust infrastructure for building, training, and deploying machine learning models.

Notable LLMs (Hosted in Bedrock)

This layer highlights large language models hosted on AWS Bedrock for generative tasks.

• Amazon Nova & Titan: Proprietary large language models developed by Amazon for generative AI applications.
• Anthropic Claude: A conversational AI model from Anthropic designed for safe and helpful interactions.
• AI21 Labs: Offers advanced language models like Jurassic for text generation tasks.
• Cohere: Provides LLMs specialized in natural language understanding and generation.
• Meta Llama: Meta’s LLM focused on efficient generative tasks.
• Stable Diffusion: A model for generating images from text prompts, hosted on AWS infrastructure.

Coding & Productivity

This layer focuses on tools to enhance coding efficiency and productivity through generative AI.

• Amazon Q (Coding and Intelligence Assist): A tool designed to assist developers with code generation, debugging, and intelligent recommendations using generative AI. Amazon Q integrates with other AWS services like QuickSight to give you business intelligence, create dashboards and more.

How I am learning Gen AI

My goal is to re-use as much as my existing experience for Gen AI, and then add additional components. If you know an area, add Gen AI components to your arsenal. For example:

• You know Serverless. Learn how Serverless can be used to do RAG, Agentic AI, and invoke LLM endpoint. Bedrock will be helpful here
• If you are into open source technology like Kubernetes, learn how you can train, and run LLMs on Kubernetes
• If you are in DevOps, learn MLOps
• If you are into storage, learn vector databases, how can they be used, how can embeddings be done

Since I am deep in Kubernetes and Serverless and meet with customers, Gen AI comes up in relation to Kubernetes and Serverless. I am learning Gen AI on Kubernetes, MLOps, and Bedrock with Serverless to answer customer questions.

ChatGPT adds a brand new feature!

ChatGPT just announced image generation and editing feature. It's pretty cool. I generated the image below using this. This is how you can too

​

Steps:

1. Select ChatGPT 4o model inside ChatGPT
2. By using the "+" sign in the chat window, upload your desired image.
3. I used prompt "turn the dog into a superhero dog and me into ghibli style cartoon". You can use the command based on your desired style and output
4. It generates the output image!

So far, Grok has been leading the image generation race. But that changes now! ChatGPT image quality is pretty dope! have you tried it yet?

Question to you readers - Have you started learning Gen AI yet? And if yes, what topics/tools/services are you learning?

If you have found this newsletter helpful, and want to support me 🙏:

Checkout my bestselling courses on AWS, System Design, Kubernetes, DevOps, and more: Max discounted links​

AWS SA Bootcamp with Live Classes, Mock Interviews, Hands-On, Resume Improvement and more: https://www.sabootcamp.com/​

Keep learning and keep rocking 🚀,

Raj