Live challenges replace theoretical tests as AWS launches three practical environments for building and securing smart agents
AWS's training and certification platform launched three new practical environments as part of the “AWS Jam” program, specifically aimed at building, deploying and securing AI agents in a live cloud environment. The move responds to the shift in software engineering toward agent-based systems, where relying on theoretical tests or multiple-choice questions is no longer sufficient to assess technical skills; engineers now need to be evaluated through hands-on work in a realistic cloud setting that mirrors production workloads.
The first track focuses on specification-driven software developmentDelivered through the Kiro development environment, the track is intended for developers, technical team leads and mid-level software operations engineers. The training spans one day and presents participants with eight technical challenges covering the specification of new features and bug-fix specifications, configuration of routing files, integration of model-context protocol servers, activation of agent hooks and construction of custom sub-agents. These challenges test a developer’s ability to connect tools to multiple cloud services such as Amazon EKS, ECS, Lambda, the Terraform server, and automated build and deployment tools.
The second track focuses on building and securing AI agents using the Amazon Bedrock Agent Core platform, offering twelve practical challenges in realistic sector-specific scenarios that span healthcare, retail, pharmacy, financial services and digital security. The challenges address runtime management, integration gateways and agent memory, enforce safety constraints through Bedrock’s protection barriers, isolate data via identity and access management, integrate tools through the model-context protocol, and include detailed training on debugging and repairing agents in the field.
The advanced track elevates the challenge to multi-agent system engineeringand its operational governance, targeting experienced developers and advanced engineering teams. The track includes twelve challenges covering direct agent-to-agent communication patterns, improved context and memory management, implementation of access-control policies based on the Cedar language, as well as monitoring and control frameworks for manufacturing, compliance and software operations sectors. These challenges rely on the AWS Standards agent development toolkit, together with monitoring and configuration-management tools such as CloudWatch, Config and Inspector, and the DynamoDB database service.
The challenge format in AWS Jam uses a one-day competitive cohort environment within a real AWS console account, rather than virtual simulations or multiple-choice-restricted tests. These experiences are offered as standalone events or attached to instructor-led training courses, where participants compete to solve open technical problems for points and to top a live leaderboard, with points deducted for requesting hints, giving teams an interactive experience that reinforces skill retention and complex problem-solving under time pressure.
The shift carries a clear operational significance for software teams and technology departments across the Gulf states, Egypt and the broader Arab region. As banks and government agencies accelerate the integration of intelligent agents into daily operations, a skills gap emerges as a major obstacle for development teams, because writing commands alone no longer suffices to build reliable systems. Data-governance and cyber-compliance requirements in regional environments impose strict limits on data movement and access privileges, making expertise in configuring Cedar policies, isolating agent identities and controlling external tool calls a critical prerequisite for protecting organizations’ digital infrastructure before moving projects to production servers.
Software competency assessment is moving in practice toward field-readiness testinginstead of certificates that measure only theoretical knowledge. In the regional job market, which sees increasing competition for AI talent, companies need to test their engineers’ ability to harden systems and uncover communication gaps between agents in real environments. Training teams to tackle open challenges in a live cloud environment gives technical leaders a practical benchmark for measuring staff skills and equips engineering teams to manage complex software solutions with operational efficiency and comprehensive security.