Amazon to let customers reshape AI models mid-training for $100K per year

Your new AI teammate never clocks out

Imagine a coworker who works through the night, mastering complex tasks without supervision. This is the reality of new frontier agents from Amazon Web Services.

AWS describes these frontier agents as a major leap from simple chatbots and says they can run autonomously for hours or days when configured to do so.

These agents handle goal-oriented projects, not just single commands. You provide a clear objective, and they devise and execute their own plan to achieve it, learning and adapting as they go.

Meet your virtual software developer

The Kiro autonomous agent is like a dedicated developer on your team. It can triage bugs, improve code, and manage tasks across multiple software repositories simultaneously. You can assign work directly from your project management tools.

Kiro can preserve context across sessions and integrate with your pull requests and feedback when you connect it to your repositories. It proposes code changes for human review, so teams retain control over what gets merged.

A security expert built into your process

The AWS Security Agent embeds deep security expertise directly into your development lifecycle. It proactively reviews design documents and scans code against your organization’s specific security rules. This happens automatically as your team builds.

It can run on-demand penetration tests and produce validated findings and recommended fixes, which help teams find and fix issues earlier in the development lifecycle when they adopt the service.

Your always available operations genius

When an app fails, the AWS DevOps Agent springs into action to find the root cause. It correlates data across your observability tools, code repositories, and deployment pipelines. This maps your system’s complex relationships to pinpoint issues fast.

AWS reports that the DevOps Agent has been used internally on thousands of escalations and provides an estimated root cause identification rate based on internal evaluations.

Agents that grow with your team

These frontier agents are built to learn and adapt over time. They connect to your team’s existing tools like Jira, GitHub, and Slack. Every code review, ticket, and decision informs their understanding of your projects.

This continuous learning makes them more valuable and context-aware as time passes. They become a shared resource that strengthens the entire team’s workflow and knowledge base.

Build your own custom AI brain

AWS also introduced a powerful tool called Nova Forge. It allows companies to build their own custom AI models by blending proprietary data with Amazon’s frontier technology.

AWS positions Nova Forge as a more cost-effective path than training a frontier model from scratch for many customers, because it allows starting from earlier checkpoints and reusing the Nova training infrastructure.

Customers like Reddit have used it to create a model specifically expert in their platform’s content. This results in AI that deeply understands unique business domains and needs.

The powerful Nova 2 model family

The intelligence behind these services comes from the new Nova 2 model family. Nova 2 Pro is a top-tier reasoning model for complex tasks like coding and long-range planning. Nova 2 Lite offers a cost-effective option for everyday workloads.

A standout, Nova 2 Omni, is a unified multimodal model. It can process text, images, video, and speech together, generating both text and images in response.

Why custom training is a game changer

Typical AI customization happens very late in a model’s development, limiting true understanding. Nova Forge allows training to start from much earlier checkpoints in the model’s development cycle. This is called open training.

This early blending of data helps the model deeply ingest domain expertise. It avoids a common problem where models forget core skills when learning new information.

From fixing problems to preventing them

The AWS DevOps Agent helps teams shift from reactive firefighting to proactive improvement. It analyzes patterns across historical incidents and system data. Then, it provides targeted recommendations to strengthen your application’s reliability.

These suggestions focus on four key areas: observability, infrastructure, deployment pipelines, and application resilience. This unlocks hidden insights in your operational data.

Real world impact and results

Companies across industries are already using these tools with dramatic results. Photography platform SmugMug uses the Security Agent for penetration tests that finish in hours, not days, at a fraction of the manual cost.

According to SmugMug and AWS, the Security Agent found a subtle business logic issue that had been missed by prior automated checks, and the agent was able to surface it through contextualized testing.

Transforming how teams work

This technology moves AI from a helpful assistant to an autonomous team member. It takes on entire multi-step projects, not just isolated tasks. This fundamental shift frees human talent to focus on high-level strategy and creative innovation.

The goal is to remove friction and administrative busy-work from the development process. It shortens the path from a great idea to a meaningful, shipped product.

Ready to see how this powerful technology stands up under pressure? Check out what happens when it takes a day off in our story about the recent Amazon and Fortnite outage.

The beginning of an agentic future

Kiro, the Security Agent, and the DevOps Agent mark the start of a new software development era. They promise to redefine what’s possible when AI acts as a true extension of your team. This future is about autonomous outcomes across the entire development lifecycle.

These tools are available in preview now. They leverage Amazon’s decades of software, security, and operational expertise to help all builders innovate with confidence.

Curious how this drive for automation is shaping the real world? See how Amazon is now putting humanoid robots to work in its warehouses for faster shipments.

Which of these AI teammates would you want on your project first? Share your pick in the comments, and give this post a thumbs up if you found it interesting.

Read More From This Brand:

• New Benefits of Amazon Prime in 2025?
• Amazon Pushes AI With Startup Mentality
• Amazon Fires First 27 Internet Satellites To Space

Don’t forget to follow us for more exclusive content on MSN.

If you liked this story, you’ll LOVE our FREE emails. Join today and be the first to get stories like this one.

This slideshow was made with AI assistance and human editing.