AWS Builder Center
Weekend Creative Agent Challenge: Serverless AI Agents on AWS for Autonomous Health Tracking

Weekend Creative Agent Challenge: Serverless AI Agents on AWS for Autonomous Health Tracking

An event-driven, serverless AI agent built on AWS that synthesizes patient data into personalized daily health digests using Amazon Bedrock, AWS Lambda, and Amazon DynamoDB.

Developer

Vision and What It Does

The core vision behind CarePulse is to build an autonomous agent that operates headlessly. Rather than requiring patients to log into an application manually every day, CarePulse executes on a schedule, synthesizes real-time patient data, generates empathetic clinical guidance, and delivers it directly to the user's inbox and web dashboard before they wake up.
CarePulse solves three key problems:
  • Medication Adherence: Automatically builds time-sensitive medication schedules and safety checklists.
  • Proactive Lifestyle Guidance: Provides tailored wellness, nutrition, and hydration goals based on active medical conditions.
  • Triage Warning Indicators: Clearly highlights specific warning signs that instruct patients when to contact a healthcare provider.

How I Built It

Building CarePulse required moving from simple prompt-response interactions to an event-driven serverless pipeline.
The build process involved setting up Amazon DynamoDB to hold patient profile state, configuring AWS Lambda as the core orchestration agent, and utilizing Amazon Bedrock via the  converse()  API for foundation model reasoning.

Key Decisions & Challenges

  • Cross-Region Model Invocation: During initial testing with  amazon.nova-micro-v1:0  via  invoke_model , standard model IDs threw validation exceptions regarding on-demand throughput. The solution was migrating to the Bedrock  converse()  API and utilizing Cross-Region Inference Profile IDs ( eu.amazon.nova-micro-v1:0 ).
  • Multi-Channel Publishing: The Lambda agent dynamically renders a responsive HTML template, uploads the live dashboard to an Amazon S3 static website bucket, and dispatches a copy directly via Amazon SES.

AWS Services Used & Architecture Overview


  • AWS EventBridge: Serves as the autonomous heartbeat, triggering the execution cron every morning at 08:00 PM SAST.

  • AWS Lambda: Hosts the Python backend that orchestrates data retrieval, AI synthesis, and publishing.

  • Amazon DynamoDB: Stores structured patient profiles, active conditions, and prescription data.

  • Amazon Bedrock: Leverages low-latency foundation models (Amazon Nova Micro) for clinical text synthesis.
  • Amazon S3: Hosts the generated public HTML dashboard.

  • Amazon SES: Delivers formatted HTML digests straight to patient inboxes.

What I Learned

Participating in the Level 200 challenge demonstrated the power of headless AI architecture. The key takeaway was that generative AI becomes far more valuable when integrated into automated serverless pipelines rather than isolated chat interfaces. Working with Amazon Bedrock’s  converse  API highlighted how effortlessly foundation models can be embedded into event-driven workflows using  boto3 .

Resources:



Any opinions in this article are those of the individual author and may not reflect the opinions of AWS.
Enjoyed reading this content? Let the author know!

Your likes, comments, shares, and saves help creators reach more builders.

Loading recommendations

Loading article