Building Practical, Cost-Efficient GenAI Solutions Using Serverless
summary
A serverless pipeline on AWS ingests files in S3, runs a Step Functions workflow that uses Textract or Transcribe to extract text, calls Bedrock or a SageMaker model for abstractive summarization, and writes summaries to DynamoDB; a demonstration processes a surgical pathology report and returns usable summaries. The implementation uses Lambda for transformations and a S3 → Step Functions → Textract/Transcribe → Bedrock/SageMaker → DynamoDB flow and shows cost cutting by bringing up and tearing down SageMaker endpoints on demand, dropping monthly inference cost from about $1,000 to $45 (~95% savings).
AI-written summary · about
chapters (12)
- 0:00Introduction
- 1:23Phases of AI
- 7:36Types of AI
- 11:56Use Case
- 15:05AI Services
- 18:06Amazon Bedrock
- 20:29Application Requirements
- 24:56Data Orchestration
- 26:58OnDemand Application
- 28:34Cost Optimization
- 30:39Making the Serverless
- 34:31Key Takeways
related talks
34mBuilding scalable end-to-end deep learning pipelines in the cloud
34mDesigning Data Intensive Applications In Serverless Architecture
39mGenerative AI Solution with AWS Serverless and Bedrock
48mBuilding Serverless AI powered Applications on AWS
47mAdvanced Serverless Architectural Patterns on AWS
41mServerless LLM with AWS Lambda and LangChain: A Revolution in Application Development
32mServerless Data Processing and Machine Learning
21mIntelligent Workflow Automation with Lambda and AI Models