This article demonstrates how to build an event-driven change data capture (CDC) pipeline that streams PostgreSQL row-level changes to Amazon S3 in near real time using AWS Fargate, EventBridge, SQS, and Lambda.

In this post, we show you how to build a fully managed, event-driven change data capture (CDC) pipeline. It streams row-level changes from Amazon RDS for PostgreSQL or Amazon Aurora PostgreSQL to Amazon S3 in near real time. You deploy the entire pipeline with a single AWS CloudFormation template, and it can run in private subnets with no internet gateway without exposing resources to the public internet.

How I would evolve a NestJS, RabbitMQ and PostgreSQL microservices project into a scalable AWS architecture, with services, trade-offs and costs.

What happens when a WhatsApp message is not just a question — but the beginning of a business transaction? A customer sends: “10 black kurtis chahiye. Size mix chalega?” A normal chatbot can generate a good reply. A real business system has to do much more.

YouTube video by Crowned Caribou

Si alguna vez trabajaste con runners autoalojados para GitHub Actions, sabes lo que implican: al principio basta con una EC2 y el agente instalado, pero después llegan el mantenimiento, los parches, el monitoreo y el costo constante, incluso sin CI corriendo. Por eso decidí hacerlo diferente: runners efímeros que nacen para ejecutar un solo job y desaparecen al terminar. Aquí te cuento cómo funciona.

Moeve standardized dbt runs across data lakes using Amazon Athena, AWS Step Functions, and AWS Fargate, reducing new project onboarding from days to 15 minutes while improving cost visibility and operational consistency.

Moeve standardized how it runs dbt across multiple data lakes by building a centralized, serverless launcher on Amazon Athena, AWS Step Functions, AWS Fargate, Amazon DynamoDB, and Amazon EventBridge, cutting new-project onboarding from days to about 15 minutes while keeping compute close to the data and orchestration loosely coupled.

Ramp migrated GPU-accelerated AI workloads from manually managed ECS on EC2 to ECS Managed Instances, eliminating fleet management overhead while maintaining operational parity with their Fargate infrastructure.

Containerization has transformed modern application development by making applications portable, scalable, and easier to manage. However, managing containers at scale can be challenging. This is where Amazon Elastic Container Service (Amazon ECS) comes in. Amazon ECS is a fully managed container orchestration service that helps developers deploy, manage, and scale containerized applications without the complexity of managing orchestration infrastructure.

AWS has published a migration guide detailing how to transition a production healthcare agent from self-managed Amazon ECS containers to the Bedrock AgentCore runtime. The guide focuses on preserving a triple-model orchestration pattern (specialist models for triage, diagnosis, and treatment planning) while eliminating container orchestration overhead. The original architecture ran on Amazon ECS with Fargate, requiring manual management of task definitions, container lifecycle, scaling policies, and custom API calls for vector retrieval from knowledge bases. In contrast, AgentCore runtime abstracts these complexities by allowing developers to define the agent through a configuration object specifying model endpoints and knowledge base connections. The runtime handles execution, state persistence, model routing, observability, and vector retrieval without requiring custom orchestration code. Key differences between the two approaches include infrastructure overhead (high in ECS vs. low in AgentCore), execution visibility (full container metrics in ECS vs. limited agent-level metrics in AgentCore), state control (custom storage in ECS vs. managed persistence in AgentCore), and deployment speed (slow container builds in ECS vs. fast configuration updates in AgentCore). The migration is recommended when infrastructure management burden outweighs the need for execution control, particularly for teams focused on agent logic rather than infrastructure. However, teams requiring fine-grained control over execution environments, complex orchestration logic, or specific compliance configurations should remain with self-managed ECS. The guide emphasizes that AgentCore is not a feature announcement but a concrete infrastructure decision, highlighting trade-offs between managed simplicity and self-hosted flexibility. AWS provides a detailed configuration example showing how to define the agent with model endpoints, knowledge base connections, and retrieval parameters. The runtime executes the configuration without requiring developers to write orchestration loops or manage retries. Vector retrieval is integrated through a built-in knowledge base connector that handles embeddings and search without custom API calls. The guide also outlines failure modes for both approaches, noting that AgentCore introduces new failure points around configuration validation and managed service dependencies while abstracting infrastructure failures. For healthcare, financial, and manufacturing agents with standard multi-model patterns, AgentCore eliminates undifferentiated infrastructure work, but teams with custom orchestration needs should maintain self-managed ECS. The migration path is presented as a practical decision point for engineering teams evaluating their infrastructure strategy.

Migrate a multi-model healthcare AI agent from self-managed Amazon ECS with AWS Fargate to Amazon Bedrock AgentCore runtime, preserving triple-model orchestration and vector-enhanced knowledge retrieval while reducing infrastructure management. The framework-agnostic pattern applies across healthcare, financial services, and manufacturing.

Migrate a multi-model healthcare AI agent from self-managed Amazon ECS with AWS Fargate to Amazon Bedrock AgentCore runtime, preserving triple-model orchestration and vector-enhanced knowledge retrieval while reducing infrastructure management. The framework-agnostic pattern applies across healthcare, financial services, and manufacturing.

AWS Lambda can be extremely cheap for unpredictable workloads—but the economics change when traffic becomes steady and high-volume. Let's use a simple cost model to understand when Fargate starts becoming financially attractive.