L37- AWS Hands-On | IAM Users, S3 Bucket, Access Keys & First Bedrock LLM Call with Boto3

Lecture 37 of the AI for Software Engineers series — Bipin Kumar starts the 3-project AWS deployment series with hands-on setup. Today's class covers cloud fundamentals, IAM user creation, S3 bucket operations, and making the first real Bedrock LLM call using Boto3 — the official AWS Python SDK. 🧠 What's Covered: 3-Project Roadmap Starting Today: Project 1 — Document Classifier (image/document type detection). Project 2 — RAG Chatbot (document question answering). Project 3 — Agentic AI Deployment (agents with MCP, memory, and APIs). All three will be deployed on AWS using Lambda, Bedrock, and Agent Core. Why Cloud? Why AWS? Running code on a local laptop means 24/7 uptime issues — power cuts, connectivity problems, and the machine must always be on. Cloud providers like AWS solve this by hosting your code in their data centres. You pay only for what you use. AWS holds the largest cloud market share. Learning AWS transfers easily to GCP or Azure — the concepts and terminology are similar. From an industry perspective: as a GenAI engineer, you do not need to be a deployment expert. DevOps and data engineers handle the infrastructure. But you need enough understanding to explain your architecture and specify what services are needed. That is exactly the level this class teaches. AWS Regions: AWS has data centres globally. Mumbai (AP-South-1) and Hyderabad (AP-South-2) are the Indian regions. Choosing a region matters for three reasons: data residency (banking and government apps must keep data inside India), cost (different regions have different prices), and model availability (not all LLM models are available in all regions). Cross-region inference: if a model you need is not available in Mumbai, AWS encrypts your data, sends it to the model's region for computation, encrypts the result, and returns it to your region. Your data is protected during transit. Root User vs IAM User: Root user is the account owner with complete control over all AWS services and billing. IAM users are created by the root user for team members with limited, specific permissions. You define which AWS services each IAM user can access. Groups allow you to apply the same set of permissions to multiple users at once — add a permission to the group and all users in that group automatically get it. Granular access principle: in a real company you only give the exact permissions needed for the project. Giving full delete access to an S3 bucket means a disgruntled employee could erase all your data. For learning, full access per service is fine. Setting Up for the 3 Projects: Services needed: S3 (file storage), Lambda (serverless compute), Bedrock (LLM and RAG), Agent Core (deploy AI agents), API Gateway (public URL for your app). These five services cover all three projects. Access Keys — Connecting Locally: After creating an IAM user, generate an access key from the security credentials section. You receive an Access Key ID and a Secret Access Key — similar to an API key but for all AWS services. Save both in a .env file on your local machine. Boto3 — The AWS SDK for Python: Boto3 is the official Python library for connecting to any AWS service. Install it once and use it throughout all three projects. You create a client object for the service you want (S3, Bedrock, Lambda) and call methods on it. S3 demo in class: created a Boto3 S3 client using the loaded access keys. Called list_buckets() and saw all existing buckets. Called create_bucket() with a new name — the bucket appeared live in the AWS console immediately. Why Boto3 instead of LangChain: LangChain is a wrapper around Boto3. Understanding Boto3 directly gives you more control and helps when debugging or when LangChain does not support a specific feature. First Bedrock LLM Call via Boto3: Create a Boto3 client for bedrock-runtime with your region. Call bedrock.converse() with the model ID, temperature, max tokens, and message list. The response format is similar to OpenAI's API. The model ID used: Claude Sonnet 4.6. Temperature set to 0.2, max tokens to 150. Output extracted from response.content. ⏭️ Next Lecture (Lecture 38): 👉 Project 1 — Document Classifier on AWS: S3 + Lambda + Bedrock end-to-end 💬 Questions about IAM setup or Boto3 errors? Drop them in the comments — Bipin replies! 📌 Subscribe so you never miss a class. #AWSHandsOn #Boto3 #IAMUser #S3Bucket #AmazonBedrock #Lambda #APIGateway #AIforEngineers #BipinKumar #CloudDeployment #GenAI #Python #AWSSetup #AccessKeys #AIInterview