Internship Report > Worklog > Week 8 Worklog

Week 8 Worklog

Week 8 Objectives:

Master Serverless Concepts: Deep dive into AWS Lambda theory, event-driven mechanisms, and the importance of Lambda Layers.
Explore Generative AI: Gain a theoretical overview of Amazon Bedrock, Foundation Models (FMs), and basic use cases.
API Architecture: Understand Amazon API Gateway theory and its critical role in Microservices architecture.
Hands-on Lab (Weekend): Successfully build an automated image processing pipeline: Upload to S3 -> Trigger Lambda (w/ Pillow Layer) -> Resize Image -> Save to destination S3 bucket.

Tasks to be carried out this week:

Day	Task	Start Date	Completion Date	Reference Material
1	AWS Lambda Core Concepts (Research) - Execution Model: Studied the Lambda lifecycle (Init, Invoke, Shutdown) and the “Cold Start” phenomenon. - Resource Model: Understood the relationship between Memory and CPU (higher RAM = proportional CPU power). - Billing: Researched pricing models: Request count + Compute duration (GB-seconds). - Concurrency: Differentiated between Reserved Concurrency (guaranteed resources) and Provisioned Concurrency (eliminated cold starts).	27/10/2025	27/10/2025	AWS Lambda Docs
2	Amazon Bedrock & GenAI Overview (Research) - Key Concepts: Explored Foundation Models (Claude, Titan, Stable Diffusion), Tokens, and Inference parameters (Temperature, Top P). - Bedrock Features: Researched theoretical concepts of Knowledge Bases (RAG) for custom data and Agents for task execution. - Security: Reviewed Bedrock’s data privacy commitments (customer data is not used to retrain AWS base models). - Use Cases: Read case studies on text summarization and image generation.	28/10/2025	28/10/2025	Bedrock User Guide
3	API Gateway Fundamentals (Research) - Architecture: Studied the role of API Gateway: Authentication, Throttling, and Caching. - Types: Compared REST API (Feature-rich) vs. HTTP API (Low-cost/Low-latency) vs. WebSocket API (Real-time). - Integration: Deep dived into Lambda Proxy Integration (passing the raw event object directly to Lambda). - Endpoint Types: Differentiated between Edge-optimized (Global), Regional (Region-specific), and Private endpoints (VPC internal).	29/10/2025	29/10/2025	API Gateway Concepts
4	S3 Event Notifications & IAM Preparation (Pre-Lab) - Event Patterns: Investigated S3 Event Notifications mechanism (specifically `s3:ObjectCreated:`) to trigger Lambda functions. - IAM Permissions:* Drafted the necessary IAM Policy: Lambda requires `s3:GetObject` (Source bucket), `s3:PutObject` (Destination bucket), and `logs:CreateLogGroup`. - Library Research: Researched the Pillow (PIL) Python library for image manipulation and the process of packaging it into a Lambda Layer (since it’s not in the standard runtime).	30/10/2025	30/10/2025	S3 Event Notifications
5	Lab: Serverless Image Resizer (Implementation) - Lambda Layer: Created and uploaded a Layer containing the `Pillow` library (compatible with Python 3.x runtime). - Coding: Developed the Lambda function (Python/Boto3) to retrieve the image from the S3 event, resize it to a thumbnail (e.g., 128x128), and save it to the target bucket. - Configuration: Configured S3 Bucket Triggers to automatically invoke the function upon file upload to the `/raw` folder. - Testing & Debug: Uploaded test images, verified results in the destination bucket, and analyzed CloudWatch Logs to resolve import/permission errors.	31/10/2025	01/11/2025	Serverless Image Resizing

Week 8 Achievements:

Solidified Serverless & AI Theory: Gained a comprehensive understanding of the Serverless ecosystem (Lambda, API Gateway) and Generative AI trends (Bedrock) through deep documentation review.
Mastered Lambda Environment: Recognized the critical importance of Lambda Layers when working with external dependencies like Pillow or Pandas, optimizing code efficiency.
Implemented Event-Driven Architecture: Successfully built a functional, real-world application: a serverless Image Processing Pipeline.
IAM & Debugging Skills: Applied “Least Privilege” principles for Lambda security and utilized CloudWatch Logs effectively for troubleshooting execution errors.