Ready to Lift Off
There was something different in the atmosphere for data scientists, AI engineers and Developers at AWS re:Invent 2025 this year with product features aligning with business problems that connect AI innovation with human intelligence. You may also read the Part 1 AWS re:Invent 2025 highlights from my Coffee n Learn newsletter at this link.
If you missed the live stream from Las Vegas you may watch the keynote below, watch innovation talks and breakout sessions from AWS Events on Youtube.
Lesson Outcome
In this lesson you will be able to identify the key announcements from the keynote delivered by VP of Agentic AI, Dr Swami Sivasubramanian and also apply your knowledge to build frontier agents.
Key Announcements
Strands Agents
The open source AI agents SDK on Amazon Bedrock removes complex orchestration and allows you to build an AI agent from your laptop with a few lines of code. Strands has already been downloaded 5 million times.
New features: AWS adds support for both Typescript and edge devices (Preview).
Production: AI agents may be deployed at scale
✓ Removes sensitive data before sharing data with other agents
✓ Includes sophisticated memory context that remembers user profiles
✓ Includes identity and access management to authenticate users and tools to manage credentials
✓ Includes modularity and enables you to observe and debug code
Amazon Bedrock – Episodic functionality in AgentCore Memory
New features from 2 December include AgentCore Policy (Preview) to create guardrails for AI agents and Evaluations (Preview) to give trust for agents in simulated scenarios.
With new feature Episodic functionality in AgentCore Memory launched on 3 December which enables AI agents to remember the past and also provides context for the current situation
Amazon SageMaker AI – AI Model Customization (GA)
New feature: Amazon Sagemaker AI – reinforcement fine-tuning
Additional model customization includes AI modelssuch as Deepseek, Amazon Nova and Qwen for reinforcement fine-tuning to reduce complexity and cost of training a reward model and providing human evaluation.
You may use natural language processing in a prompt to create a use case, the AI model (teacher model) will adjust the base model (student model).
Amazon Nova Forge (GA)
Organizations may build a frontier model with Amazon Nova curated training data, the Nova AI model may be tailored for your industry use case such as healthcare whilst reducing cost.
Amazon Nova Act (GA)
You may build web browser automation for your tasks. The state will be consistent for both training and production.
With reinforcement learning gyms, the AI agents will take action and learn from outcomes.
You may try the Amazon Nova Act playground to experience the UI workflow automation.
What is a frontier agent?
At AWS re:Invent 2025, a new class of AI agents was introduced called frontier agents which are autonomous and can continue working for long periods of time and are scalable to your needs.
AWS introduced three frontier agents:
- Kiro autonomous agent(Preview)
Kiro autonomous agent maintains context and assists teams with software development tasks and keeps building knowledge from your interactions.
For example Kiro may help you to update python dependencies from and older version to your latest version and also assist with other software tasks whilst maintaining context aware knowledge so that you do not need to provide new knowledge for each session. Kiro will try to solve the problem on their own without your intervention.
- AWS Security agent (Preview)
This is a frontier agent that can provide security of your applications throughout a development lifecycle.
The virtual security agent can perform tasks such as penetration testing and identify security risks.
What are the features?
- You may design a security overview
- Analyzes your pull requests against your company’s security requirements
- Provides on-demand penetration testing
What are the benefits?
- Validates your security requirements
- Provides on-demand access
- Scalable security reviews
- Provides actionable pull request to remediate security risks
- Multi and hybrid cloud support
In this public preview you may access AWS Security agent in the US East (N. Virginia) Region.
3.AWS DevOps agent
This is an AWS frontier agent that helps you to prevent and respond to production incidents.
AWS DevOps agent is already used by Commonwealth Bank of Australia.
What are the features?
The AWS DevOps agent integrates with your existing tools to support your workflows.
You may learn more about configuration from the AWS User Guide.
What are the benefits?
- Responds to production incidents quickly
- Prevents future incidents
- Provides insights into your workloads
Next Month – Start your 2026 with momentum
AWS Community Builders Program 2026 – Application Opens in January 2026
If you are passionate about content creation and being part of a vibrant global AWS Community you may sign up for an AWS Builders ID and register for AWS Builders Center before January 2026 to receive all the latest information for your learning.
You may also place your name on the waitlistto be notified as soon as the application opens.
Join an AWS User Group
If you would like to learn data, AI, cloud, serverless and more – you may join an AWS User Groupin your city. You may join the AWS Builder Center and locate a user group who will support you as you build new skills and participate in AWS Summits and AWS Community days.
Join a Cloud Club
If you are a university student, you may also sign up for an AWS Builders ID and locate a Cloud Club on campus in your city to support you as you learn cloud, data and AI skills.
Until the next update, happy learning! 😀
Wishing everyone happy holidays!