Change
68be6df
68be6dfa4e4b2967a84146dc3600060247756e5a · commit on GitHub
aws-blog-feed: changed (622382 bytes, HTTP 200)
raw/aws-blog-feed/response.xml added
- Source
- aws-blog-feed
- Lines added
- +5,284
- Lines removed
- -0
- Stored bytes at this commit
- 622,382
- Timestamp
- observed
- Raw artifact at this commit
- raw/aws-blog-feed/response.xml
Recorded headers
| observed_at | 2026-09-01T23:17:48.662Z |
|---|---|
| origin_date | null |
| status | 200 |
| final URL | https://aws.amazon.com/blogs/machine-learning/feed/ |
| etag | null |
| last-modified | Tue, 01 Sep 2026 23:02:05 GMT |
| date | Tue, 01 Sep 2026 23:17:48 GMT |
| age | null |
| cache-control | null |
| cf-cache-status | null |
| content-encoding | null |
| content-length | null |
@
@@ -0,0 +1,5284 @@+
<?xml version="1.0" encoding="UTF-8" standalone="no"?><rss xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:slash="http://purl.org/rss/1.0/modules/slash/" xmlns:sy="http://purl.org/rss/1.0/modules/synd…+
+
<channel>+
<title>Artificial Intelligence</title>+
<atom:link href="https://aws.amazon.com/blogs/machine-learning/feed/" rel="self" type="application/rss+xml"/>+
<link>https://aws.amazon.com/blogs/machine-learning/</link>+
<description>Official Machine Learning Blog of Amazon Web Services</description>+
<lastBuildDate>Tue, 01 Sep 2026 19:13:35 +0000</lastBuildDate>+
<language>en-US</language>+
<sy:updatePeriod>+
hourly </sy:updatePeriod>+
<sy:updateFrequency>+
1 </sy:updateFrequency>+
+
<item>+
<title>Introducing Claude Fable 5.1 on AWS</title>+
<link>https://aws.amazon.com/blogs/machine-learning/introducing-claude-fable-5-1-on-aws/</link>+
+
<dc:creator><![CDATA[Dani Mitchell]]></dc:creator>+
<pubDate>Tue, 01 Sep 2026 19:12:43 +0000</pubDate>+
<category><![CDATA[Amazon Bedrock]]></category>+
<category><![CDATA[Announcements]]></category>+
<category><![CDATA[Intermediate (200)]]></category>+
<guid isPermaLink="false">fe5a761d13fdea1cd6343d646c1e042d79f66815</guid>+
+
<description>Claude Fable 5.1 is now available on Amazon Bedrock and Claude Platform on AWS. This post covers Claude Fable 5.1's improvements, the Enterprise Frontier Safeguards for keeping your data in a cloud environment you control, and how to start building with the model on Amazon Bedrock.…+
<content:encoded><p>Today, we’re excited to announce the availability of Claude Fable 5.1 on Amazon Bedrock and Claude Platform on AWS. Claude Fable 5.1 delivers frontier intelligence for ambitious tasks across coding, scientific research, and enterprise workflows.</p> +
<p>Given its capabilities, Anthropic has designated Fable 5.1 a <em>Covered Model</em>, a category of Claude models that carry additional data retention, safety review, and access policies wherever they’re offered. For more information, see Amazon Bedrock <a href="https://docs.a…+
<p>This post covers Claude Fable 5.1’s improvements, the Enterprise Frontier Safeguards, and how to start building with the model on Amazon Bedrock.</p> +
<h2 id="what-makes-claude-fable-5.1-different">What makes Claude Fable 5.1 different</h2> +
<p>Anthropic reports Fable 5.1 is a clear improvement over Fable 5 on the hardest reasoning tests Anthropic runs. These tests include competition mathematics, graduate-level questions in engineering and the sciences, and long multi-step problems where reliability is key. In day-to-day use that…+
<ul> +
<li><strong>Agentic coding.</strong> Carries more of a project on its own, from code base-spanning features to code review and performance work, across multi-hour sessions. It is also more honest: if it gets stuck it says so, and it is less likely to disable a failing test to pass…+
<li><strong>Autonomous operation.</strong> Built for multi-hour jobs that span many applications. It plans, uses the tools it needs, recovers when a step fails, and keeps you updated without being asked.</li> +
<li><strong>End-to-end knowledge work.</strong> Takes an analysis from first question to finished document, doing the research, building the spreadsheet, writing the memo or deck, and checking its numbers as it goes. Built for everyday finance, accounting, and healthcare work.<…+
<li><strong>Scientific research.</strong> Supports research campaigns from literature and hypotheses to models, experiments, and formal verification.</li> +
<li><strong>Improved usability.</strong> Keeps you updated on long tasks, writes more clearly, and follows instructions more closely.</li> +
</ul> +
<h2 id="data-retention">Data retention</h2> +
<p>Because Claude Fable 5.1 is a <a href="https://support.claude.com/en/articles/15425695-covered-models" target="_blank" rel="noopener">Covered Model</a>, its use is subject to data retention for up to 30 days and human review by Amazon personnel. With Amazon Bedrock, you can cont…+
<h2 id="enterprise-frontier-safeguards">Enterprise Frontier Safeguards</h2> +
<p><a href="https://www.anthropic.com/news/enterprise-frontier-safeguards" target="_blank" rel="noopener">Enterprise Frontier Safeguards</a> (EFS), built in partnership between AWS and Anthropic, will help eligible customers use Claude Fable 5 and Claude Fable 5.1 models while keep…+
<p>If you are an EFS-eligible customer, you can use Claude Fable 5 and Claude Fable 5.1 with zero data retention (ZDR) on Amazon Bedrock and Claude Platform on AWS. This is available for internal use through December 31, 2026. Additional Enterprise Frontier Safeguards will be available later t…+
<h2 id="getting-started-with-claude-fable-5.1-on-amazon-bedrock">Getting started with Claude Fable 5.1 on Amazon Bedrock</h2> +
<p>To try Fable 5.1, open the <a href="https://console.aws.amazon.com/bedrock/" target="_blank" rel="noopener">Amazon Bedrock console</a>, go to <strong>Test &gt; Playground</strong>, and select Fable 5.1 as the model. From there, you can run a prompt directly again…+
<p>Programmatically, you can call the model with the <a href="https://docs.aws.amazon.com/bedrock/latest/userguide/model-parameters-anthropic-claude-messages.html" target="_blank" rel="noopener">Anthropic Messages API</a> against <code>bedrock-runtime</code> (through th…+
<h3 id="prerequisites">Prerequisites</h3> +
<ol type="1"> +
<li>Active AWS account with Amazon Bedrock access.</li> +
<li>AWS Command Line Interface (AWS CLI) installed and configured.</li> +
<li>Python 3.10+.</li> +
<li>Boto3 installed: <code>pip install boto3</code>.</li> +
<li>AWS Identity and Access Management (IAM) permissions: <code>bedrock:InvokeModel</code>and <code>bedrock:InvokeModelWithResponseStream</code>.</li> +
</ol> +
<p>Here’s a quick example using the AWS SDK for Python (Boto3):</p> +
<div class="hide-language"> +
<pre><code class="language-python">import boto3+
import json+
+
# Create a Bedrock Runtime client+
bedrock_runtime = boto3.client(+
service_name="bedrock-runtime",+
region_name="us-east-1"+
)+
+
# Invoke Claude Fable 5.1+
response = bedrock_runtime.invoke_model(+
modelId="global.anthropic.claude-fable-5-1",+
contentType="application/json",+
accept="application/json",+
body=json.dumps({+
"anthropic_version": "bedrock-2023-05-31",+
"max_tokens": 4096,+
"messages": [+
{+
"role": "user",+
"content": "An S3 bucket serves 40 TB/month egress. Estimate the monthly egress cost at $0.09/GB, and state one architecture change to cut it. Show the calculation, keep it under 120 words."+
}+
]+
})+
)+
+
result = json.loads(response["body"].read())+
# Fable 5.1 is a reasoning model: the response may include a thinking block+
# before the text block, so select the text block rather than a fixed index.+
print(next(b["text"] for b in result["content"] if b["type"] == "text"))</code></pre> +
</div> +
<p>You can explore the <a href="https://github.com/aws-samples/anthropic-on-aws/blob/main/notebooks/claude_fable_5_1_getting_started/claude-fable-5-1-getting-started.ipynb" target="_blank" rel="noopener">Getting Started notebook</a> for more examples.</p> +
<h2 id="availability">Availability</h2> +
<p>Claude Fable 5.1 is available today on Amazon Bedrock through the US Geo CRIS (us.) and Global CRIS (global.) inference profiles. In AWS GovCloud (US), it is available on both the bedrock-runtime and bedrock-mantle endpoints. See the <a href="https://docs.aws.amazon.com/bedrock/latest/us…+
<p>Give Claude Fable 5.1 a try in the <a href="https://console.aws.amazon.com/bedrock" target="_blank" rel="noopener">Amazon Bedrock console</a>, in <a href="https://console.aws.amazon.com/claude-platform/" target="_blank" rel="noopener">Claude Platform on AWS</a>, or e…+
<hr style="width: 100%"> +
<h2>About the authors</h2> +
<footer> +
<div class="blog-author-box"> +
<div class="blog-author-image"> +
<p><img class="alignleft size-full" src="https://d2908q01vomqb2.cloudfront.net/f1f836cb4ea6efb2a0b1b99f41ad8b103eff4b59/2026/05/11/ml-20969-image-6.png" alt="Dani Mitchell" width="100" height="100"></p> +
</div> +
<h3 class="lb-h4">Dani Mitchell</h3> +
<p>Dani is a Sr. generative AI Specialist Solutions Architect at AWS and the SA lead for Amazon Bedrock Knowledge Bases. He helps enterprises across the world design and deploy generative AI solutions using Amazon Bedrock and Anthropic’s models and capabilities to build scalable, production-…+
</div> +
<div class="blog-author-box"> +
<div class="blog-author-image"> +
<p><img loading="lazy" class="alignleft size-full" src="https://d2908q01vomqb2.cloudfront.net/f1f836cb4ea6efb2a0b1b99f41ad8b103eff4b59/2026/07/24/ML-21458-3.png" alt="Aamna Najmi" width="100" height="100"></p> +
</div> +
<h3 class="lb-h4">Aamna Najmi</h3> +
<p>Aamna is a Senior Specialist Solutions Architect for Generative AI focusing on Anthropic models and operationalizing and governing generative AI systems at scale on Amazon Bedrock. She helps ISVs solve their challenges, embrace innovation, and create new business opportunities with Amazon…+
</div> +
<div class="blog-author-box"> +
<div class="blog-author-image"> +
<p><img loading="lazy" class="alignleft size-full" src="https://d2908q01vomqb2.cloudfront.net/f1f836cb4ea6efb2a0b1b99f41ad8b103eff4b59/2026/04/21/ml-20855-image-2.png" alt="Sofian Hamiti" width="100" height="100"></p> +
</div> +
<h3 class="lb-h4">Sofian Hamiti</h3> +
<p>Sofian is a technology leader with over 12 years of experience building AI solutions, and leading high-performing teams to maximize customer outcomes. He is passionate about empowering diverse talents to drive global impact and achieve their career aspirations.</p> +
</div> +
<div class="blog-author-box"> +
<div class="blog-author-image"> +
<p><img loading="lazy" class="alignleft size-full" src="https://d2908q01vomqb2.cloudfront.net/f1f836cb4ea6efb2a0b1b99f41ad8b103eff4b59/2026/08/31/ML-21797-1.jpg" alt="Antonio Rodriguez" width="100" height="100"></p> +
</div> +
<h3 class="lb-h4">Antonio Rodriguez</h3> +
<p>Antonio is a Principal Generative AI Tech Leader at Amazon Web Services. He helps companies of all sizes solve their challenges, embrace innovation, and create new business opportunities with Amazon Bedrock.</p> +
</div> +
<div class="blog-author-box"> +
<div class="blog-author-image"> +
<p><img loading="lazy" class="alignleft size-full" src="https://d2908q01vomqb2.cloudfront.net/f1f836cb4ea6efb2a0b1b99f41ad8b103eff4b59/2026/08/31/ML-21797-2.jpg" alt="Ayan Ray" width="100" height="100"></p> +
</div> +
<h3 class="lb-h4">Ayan Ray</h3> +
<p>Ayan is a Principal Partner Solutions Architect and AI Tech Lead at AWS, serving as the Worldwide Tech Lead for Anthropic at AWS. He works at the intersection of cloud architecture and Artificial Intelligence, helping organizations adopt and scale Anthropic’s technologies on AWS.</p>…+
</div> +
</footer></content:encoded>+
+
+
+
</item>+
<item>+
<title>From theory to delivery: How Atos upskilled 400 engineers in agentic AI</title>+
<link>https://aws.amazon.com/blogs/machine-learning/from-theory-to-delivery-how-atos-upskilled-400-engineers-in-agentic-ai/</link>+
+
<dc:creator><![CDATA[Rajesh Babu Nuvvula]]></dc:creator>+
<pubDate>Tue, 01 Sep 2026 16:17:54 +0000</pubDate>+
<category><![CDATA[Amazon Bedrock]]></category>+
<category><![CDATA[Amazon SageMaker]]></category>+
<category><![CDATA[Customer Solutions]]></category>+
<guid isPermaLink="false">c09782028dd4cf40d9245e4cd28716a51318e5b0</guid>+
+
<description>When Atos set out to upskill 400 engineers in agentic AI, hands-on learning was the missing ingredient. Over three days, engineers built multi-agent systems on AWS through an AI League event. This post explains why Atos chose the format, what engineers built and learned, and what o…+
<content:encoded><p>When Atos set out to upskill 400 engineers from theory to delivery in agentic AI, the team faced a familiar challenge: how to build real-world capability, not only theoretical knowledge. Online courses and classroom-based instruction build foundations, but they do…+
<p>Through the <a href="https://aws.amazon.com/partners/atos/" target="_blank" rel="noopener">Atos partnership with AWS</a>, we had already seen that hands-on learning was the missing ingredient in effective AI enablement. We had previously delivered practical upskilling in reinfor…+
<p>In 2026, Atos partnered with AWS to run an agentic AI League event for 400 engineers. Over three days, engineers moved from limited hands-on experience to building multi-agent systems with pathfinding, guardrails, memory, and fine-tuned models. They competed on a live leaderboard that score…+
<p>Participant skill levels varied widely. Some were developers with existing AWS experience. Others were using AWS for the first time or held less technical roles such as product owners and project managers.</p> +
<ul> +
<li>5% had no prior knowledge of agentic AI.</li> +
<li>25% had basic awareness of the topic.</li> +
<li>50% understood the topic but had no hands-on experience.</li> +
<li>20% had practical experience with agentic AI services.</li> +
</ul> +
<p>This post explains why we chose the AI League format, what engineers built and learned, which AWS services were involved, and what other enterprises should consider when running a similar event.</p> +
<h2 id="why-the-aws-ai-league">Why the AWS AI League?</h2> +
<p>Atos has a strategic commitment to agentic AI, including the development of <a href="https://atos.net/en/services/ai-atos-sovereign-agentic-studios" target="_blank" rel="noopener">Sovereign Agentic AI Studios</a> in multiple locations worldwide. We needed a way to upskill our en…+
<p>The AI League format offered several advantages over conventional training:</p> +
<ul> +
<li>Immediate practical application – engineers built working agentic systems, rather than simply reading about them.</li> +
<li>Competitive motivation – the leaderboard created urgency and engagement that passive learning rarely achieves.</li> +
<li>Real AWS services – everything built during the event used native AWS services that engineers could apply in client delivery, including <a href="https://aws.amazon.com/bedrock/" target="_blank" rel="noopener">Amazon Bedrock</a>, <a href="https://aws.amazon.com/bedrock/agent…+
<li>Measurable outcomes – AI League leaderboard scores reflected functional completeness and solution efficiency.</li> +
</ul> +
<h2 id="time-commitment-and-format">Time commitment and format</h2> +
<p>AWS AI League is a turnkey solution, so setup was straightforward. Setup required only a few calls with AWS to agree on logistics and event details, plus a mechanism to advertise the event and collect participant information.</p> +
<p>Because AWS AI League was delivered through the <a href="https://workshops.aws/" target="_blank" rel="noopener">AWS Workshop Studio</a>, the event could be run over one to three days. To give engineers maximum flexibility, we chose a three-day format. Participant time commitment…+
<ul> +
<li>Initial kick-off workshop (2 hours) – introduced the AI League, the AWS services used, and the format of the challenges.</li> +
<li>Daily office hours call (1 hour) – provided support for engineers who needed help or wanted to share ideas.</li> +
<li>Top three finale (1 hour) – where we crowned our 2026 champion.</li> +
</ul> +
<p>Outside these scheduled sessions, engineers were free to iterate on their agentic AI solutions around their existing commitments.</p> +
<h2 id="the-ai-league-challenge">The AI League challenge</h2> +
<p>Engineers built an autonomous AI agent that navigated a dungeon maze. The agent had to find a path through the map, solve challenges on various tiles, avoid traps, and reach the treasure. All of this had to be completed within a time limit and with limited lives.</p> +
<p>The following figure shows an overview of an AWS AI League map.</p> +
<div style="width: 810px" class="wp-caption alignnone">+
<a href="https://d2908q01vomqb2.cloudfront.net/f1f836cb4ea6efb2a0b1b99f41ad8b103eff4b59/2026/08/28/ML-21185-1.png" target="_blank" rel="noopener"><img src="https://d2908q01vomqb2.cloudfront.net/f1f836cb4ea6efb2a0b1b99f41ad8b103eff4b59/2026/08/28/ML-21185-1.png" alt="Screenshot of the AWS A…+
<p class="wp-caption-text">Figure 1: Overview of an AWS AI League map</p>+
</div> +
<p>The scoring model rewarded:</p> +
<ul> +
<li>Successful challenge completion – correct answers earned points, while incorrect answers cost lives.</li> +
<li>Coin collection – unlike challenges, coins carried no risk and required no additional time to collect.</li> +
<li>Map completion – engineers received a treasure bonus for reaching the treasure within the allotted time.</li> +
<li>Life retention – each remaining life at the end of the challenge earned additional points.</li> +
<li>Efficiency – concise agent responses outscored more verbose ones.</li> +
<li>Fine-tuning – developing specialist small language models earned bonus points.</li> +
</ul> +
<p>The challenge types tested different AI engineering skills:</p> +
<table border="1px" width="100%" cellpadding="10px"> +
<tbody> +
<tr> +
<td><strong>Challenge</strong></td> +
<td><strong>Skill tested</strong></td> +
<td><strong>AWS Service(s)</strong></td> +
</tr> +
<tr> +
<td>Violent Violet</td> +
<td>AI safety and content filtering</td> +
<td>Amazon Bedrock Guardrails</td> +
</tr> +
<tr> +
<td>Blue Brain</td> +
<td>Code generation and execution</td> +
<td> <p>AWS Lambda,</p> <p>AgentCore Code Interpreter, a capability of Amazon Bedrock AgentCore</p></td> +
</tr> +
<tr> +
<td>Memento</td> +
<td>Context retention across interactions</td> +
<td>AgentCore memory, a capability of Amazon Bedrock AgentCore</td> +
</tr> +
<tr> +
<td>Dark Prophet</td> +
<td>Information retrieval from web sources</td> +
<td> <p>AWS Lambda,</p> <p>AgentCore Code Interpreter</p></td> +
</tr> +
<tr> +
<td>Bonehead</td> +
<td>General knowledge with token efficiency</td> +
<td>Amazon Bedrock (prompt engineering)</td> +
</tr> +
<tr> +
<td>Healthcare API</td> +
<td>Structured data extraction</td> +
<td>Amazon Bedrock (prompt engineering)</td> +
</tr> +
<tr> +
<td>Keys &amp; Doors</td> +
<td>Context retention across interactions</td> +
<td>AgentCore memory</td> +
</tr> +
<tr> +
<td>Spikes &amp; Coins</td> +
<td>Pathfinding and risk assessment</td> +
<td>AWS Lambda</td> +
</tr> +
</tbody> +
</table> +
<p>The following diagram shows the solution architecture.</p> +
<div style="width: 810px" class="wp-caption alignnone">+
<a href="https://d2908q01vomqb2.cloudfront.net/f1f836cb4ea6efb2a0b1b99f41ad8b103eff4b59/2026/08/28/ML-21185-2.jpg" target="_blank" rel="noopener"><img src="https://d2908q01vomqb2.cloudfront.net/f1f836cb4ea6efb2a0b1b99f41ad8b103eff4b59/2026/08/28/ML-21185-2.jpg" alt="Solution architecture o…+
<p class="wp-caption-text">Figure 2: Overview of the AWS AI League architecture</p>+
</div> +
<h3 id="amazon-bedrock">Amazon Bedrock</h3> +
<p>With Amazon Bedrock, engineers accessed the models that powered the agent’s reasoning. For model availability by Region, refer to <a href="https://docs.aws.amazon.com/bedrock/latest/userguide/models-regions.html" target="_blank" rel="noopener">Supported models by AWS Region in Amazon …+
<ul> +
<li>Select appropriate models for different tasks, beginning with a well-known model and then performing inference against their own fine-tuned model.</li> +
<li>Engineer effective system prompts to answer questions efficiently or delegate to sub-agents and tools.</li> +
<li>Manage token usage and cost.</li> +
</ul> +
<h3 id="amazon-bedrock-agentcore">Amazon Bedrock AgentCore</h3> +
<p>With Amazon Bedrock AgentCore, a platform to build, connect, and optimize agents at scale with any framework or model, engineers orchestrated multi-agent systems. Engineers used the following AgentCore capabilities during the AI League:</p> +
<ul> +
<li>AgentCore runtime, a capability of Amazon Bedrock AgentCore – engineers hosted their agent containers, which processed challenge tiles and returned scored responses within the time limit.</li> +
<li>AgentCore Gateway, a capability of Amazon Bedrock AgentCore – engineers routed tool calls from agents to Lambda functions through the Model Context Protocol (MCP) for pathfinding, web scraping, and code execution.</li> +
<li>AgentCore memory, a capability of Amazon Bedrock AgentCore – engineers persisted context across interactions so agents could recall previous events like collected keys and solved challenges.</li> +
<li>AgentCore Code Interpreter, a capability of Amazon Bedrock AgentCore – engineers executed code securely in an isolated sandbox for computational challenges.</li> +
</ul> +
<h3 id="amazon-bedrock-guardrails">Amazon Bedrock Guardrails</h3> +
<p>With Amazon Bedrock Guardrails, you can filter content to protect against harmful inputs and outputs. Engineers configured:</p> +
<ul> +
<li>Denied topics – specific content that had to be blocked.</li> +
<li>Content filters – thresholds for hate, violence, and misconduct.</li> +
<li>Input and output blocking with custom messages.</li> +
</ul> +
<h3 id="aws-lambda">AWS Lambda</h3> +
<p>With AWS Lambda, you can build custom tool functions for tasks that models cannot reliably handle on their own:</p> +
<ul> +
<li>Pathfinding – navigating the map using algorithms such as Breadth-First Search (BFS).</li> +
<li>Code Interpreter – executing code for computational challenges.</li> +
<li>Web scraping – fetching and parsing web pages for information retrieval.</li> +
</ul> +
<h3 id="amazon-sagemaker">Amazon SageMaker</h3> +
<p>With Amazon SageMaker, you can build your development environment and fine-tune models. Engineers used Reinforcement Learning from Verifiable Rewards (RLVR):</p> +
<ul> +
<li>Amazon SageMaker Studio provided the development environment, including an integrated development environment (IDE) with built-in AI development tools.</li> +
<li>Serverless fine-tuning trained custom models on participant-created datasets.</li> +
<li>Fine-tuned models were then deployed to inference endpoints to serve traffic.</li> +
</ul> +
<p>The following figure shows the fine-tuning model workflow.</p> +
<div id="attachment_138466" style="width: 2764px" class="wp-caption alignleft">+
<a href="https://d2908q01vomqb2.cloudfront.net/f1f836cb4ea6efb2a0b1b99f41ad8b103eff4b59/2026/09/01/Screenshot-2026-09-01-at-10.57.33 AM.png"><img aria-describedby="caption-attachment-138466" loading="lazy" class="size-full wp-image-138466" src="https://d2908q01vomqb2.cloudfront.net/f1f836c…+
<p id="caption-attachment-138466" class="wp-caption-text">Figure 3: Fine-tuning model overview</p>+
</div> +
<h2 id="what-our-engineers-learned">What our engineers learned</h2> +
<p>The AWS AI League surfaced several practical engineering lessons that are directly applicable to production agentic AI work.</p> +
<h3 id="prompt-engineering-under-constraints">Prompt engineering under constraints</h3> +
<p>One of the clearest lessons was that success did not come from writing prompts alone, but from writing prompts that worked under real constraints. Every extra token cost points. Every unnecessary tool call consumed time and cost points. Engineers quickly discovered that getting a solution t…+
<h3 id="multi-agent-architecture-decisions">Multi-agent architecture decisions</h3> +
<p>Engineers had to decide how many agents to include in their architecture, from single-purpose agents with specialist tools to more multifunctional agents. Each approach involved trade-offs in token usage, latency, and reliability. This closely reflects real production decisions about agent …+
<p>The following figure shows how different agentic systems answered challenges during the finale.</p> +
<div style="width: 810px" class="wp-caption alignnone">+
<a href="https://d2908q01vomqb2.cloudfront.net/f1f836cb4ea6efb2a0b1b99f41ad8b103eff4b59/2026/08/28/ML-21185-4.png" target="_blank" rel="noopener"><img src="https://d2908q01vomqb2.cloudfront.net/f1f836cb4ea6efb2a0b1b99f41ad8b103eff4b59/2026/08/28/ML-21185-4.png" alt="Screenshot of the AWS A…+
<p class="wp-caption-text">Figure 4: Overview of different agentic systems answering challenges</p>+
</div> +
<h3 id="guardrail-configuration">Guardrail configuration</h3> +
<p>The Violent Violet challenge highlighted an important lesson: guardrails had to be precise enough to block undesirable content, without over-blocking legitimate queries. If they were too aggressive, engineers failed other challenges. If they were too permissive, they failed the guardrail ch…+
<h3 id="pathfinding-algorithm-design">Pathfinding algorithm design</h3> +
<p>Building the pathfinding tool required engineers to think about:</p> +
<ul> +
<li>Multiple strategies for different scenarios (speed versus score maximization).</li> +
<li>Risk assessment (spikes cost lives, while walls end the game).</li> +
<li>Time budgeting (should we visit every challenge, or go straight for the treasure?)</li> +
<li>Dependency ordering (for example, collecting the key before attempting the door).</li> +
</ul> +
<p>The following figure shows examples of multiple pathfinding strategies on the same map.</p> +
<div style="width: 810px" class="wp-caption alignnone">+
<a href="https://d2908q01vomqb2.cloudfront.net/f1f836cb4ea6efb2a0b1b99f41ad8b103eff4b59/2026/08/28/ML-21185-5.png" target="_blank" rel="noopener"><img src="https://d2908q01vomqb2.cloudfront.net/f1f836cb4ea6efb2a0b1b99f41ad8b103eff4b59/2026/08/28/ML-21185-5.png" alt="Screenshot of the AWS A…+
<p class="wp-caption-text">Figure 5: Example of multiple different pathfinding strategies</p>+
</div> +
<h3 id="the-value-of-observability">The value of observability</h3> +
<p>Engineers who checked their Amazon CloudWatch Logs between runs appeared to improve more quickly. Those who guessed what had gone wrong often made slower progress. This reinforced a core engineering principle: instrument the solution properly and observe before acting.</p> +
<p>The following figure shows how to troubleshoot a Lambda function using Amazon CloudWatch Logs.</p> +
<div style="width: 810px" class="wp-caption alignnone">+
<a href="https://d2908q01vomqb2.cloudfront.net/f1f836cb4ea6efb2a0b1b99f41ad8b103eff4b59/2026/08/28/ML-21185-6.png" target="_blank" rel="noopener"><img src="https://d2908q01vomqb2.cloudfront.net/f1f836cb4ea6efb2a0b1b99f41ad8b103eff4b59/2026/08/28/ML-21185-6.png" alt="Screenshot of Amazon Cl…+
<p class="wp-caption-text">Figure 6: Troubleshooting a Lambda function with Amazon CloudWatch Logs</p>+
</div> +
<h3 id="using-ai-to-build-agentic-ai-solutions">Using AI to build agentic AI solutions</h3> +
<p>Engineers who used AI developer tools such as Kiro often made rapid progress. Those who shared the full context of the challenge with their AI tools tended to achieve better results more quickly. This reinforces a broader lesson: AI tools deliver more value when grounded in the specific pro…+
<h2 id="results-and-outcomes">Results and outcomes</h2> +
<p>The event delivered:</p> +
<ul> +
<li>Based on the event registration data, 400 engineers gained hands-on experience in agentic AI.</li> +
<li>Engineers developed practical skills in Amazon Bedrock, Amazon Bedrock AgentCore, AWS Lambda, Amazon Bedrock Guardrails, Kiro, and Amazon SageMaker.</li> +
<li>Internal champions emerged with greater confidence in applying agentic AI in client engagements.</li> +
<li>The competitive format encouraged knowledge sharing and helped break down barriers between teams.</li> +
</ul> +
<p>The top-performing solutions demonstrated thoughtful engineering. They included custom pathfinding strategies, carefully tuned guardrails, memory-aware agents, and fine-tuned models designed to reduce token usage. Congratulations to our top three performers: James Ponter, Adam Różewicki, an…+
<blockquote>+
<p><em>“Academic learning gives you the foundation, but the AWS AI League puts it under pressure in a way that genuinely changes how you think. Building a production-style multi-agent architecture on real AWS infrastructure, not a toy project, but something scored on both performance an…+
</blockquote> +
<p>The following figure shows the finale leaderboard.</p> +
<div style="width: 810px" class="wp-caption alignnone">+
<a href="https://d2908q01vomqb2.cloudfront.net/f1f836cb4ea6efb2a0b1b99f41ad8b103eff4b59/2026/08/28/ML-21185-7.png" target="_blank" rel="noopener"><img src="https://d2908q01vomqb2.cloudfront.net/f1f836cb4ea6efb2a0b1b99f41ad8b103eff4b59/2026/08/28/ML-21185-7.png" alt="Screenshot of the AWS A…+
<p class="wp-caption-text">Figure 7: Our finale leaderboard</p>+
</div> +
<p>A closing comment from Chris Byrne, Global Head of AWS Alliance at Atos, on the benefits of the gamified learning approach taken to AWS upskilling:</p> +
<blockquote>+
<p><em>“Taking the step from theoretical knowledge to hands-on experience can be daunting on the one hand, and challenging knowing where and how to start on the other. Atos has successfully used the AWS leagues for Reinforcement learning with AWS DeepRacer, model fine-tuning with AI Lea…+
</blockquote> +
<h2 id="getting-started">Getting started</h2> +
<p>AWS AI League is available for enterprise events throughout 2026, and at select AWS Summits and virtual events. The format is flexible, ranging from half-day workshops to multi-day hackathons, and AWS provides the infrastructure, accounts, and facilitation support.</p> +
<p>To explore running an AI League event for your organization:</p> +
<ul> +
<li>Visit the <a href="https://aws.amazon.com/ai/aileague/" target="_blank" rel="noopener">AWS AI League page</a> to learn about the program.</li> +
<li>Contact your AWS account team to discuss a private event.</li> +
<li>Join the <a href="https://builder.aws.com/connect/space/7e5f51ef-0919-32da-aaa7-ddf263651d69/aws-ai-league" target="_blank" rel="noopener">AWS AI League Builder Space</a> for official announcements and the <a href="https://builder.aws.com/connect/space/7148b02a-ef8c-3a67-97…+
</ul> +
<p>AWS AI League is more than a competition. It is an effective way to accelerate practical AI skills development and encourage idea sharing. To learn more or explore running an event for your organization, visit the <a href="https://aws.amazon.com/ai/aileague/" target="_blank" rel="noopene…+
<ul> +
<li>Learn more about the AWS AI League’s agentic AI and model customization challenges in <a href="https://aws.amazon.com/blogs/machine-learning/aws-ai-league-model-customization-and-agentic-showdown/" target="_blank" rel="noopener">AWS AI League: Model customization and agentic showdow…+
<li>See how a student went from beginner to champion in <a href="https://aws.amazon.com/blogs/machine-learning/from-beginner-to-champion-a-students-journey-through-the-aws-ai-league-asean-finals/" target="_blank" rel="noopener">From beginner to champion: A student’s journey through the …+
<li>For guidance on taking agentic AI from competition to production, see <a href="https://aws.amazon.com/blogs/machine-learning/ai-agents-in-enterprises-best-practices-with-amazon-bedrock-agentcore/" target="_blank" rel="noopener">AI agents in enterprises: Best practices with Amazon Be…+
</ul> +
<hr style="width: 100%"> +
<h2>About the authors</h2> +
<footer> +
<div class="blog-author-box"> +
<div class="blog-author-image"> +
<p><img loading="lazy" class="alignleft size-full" src="https://d2908q01vomqb2.cloudfront.net/f1f836cb4ea6efb2a0b1b99f41ad8b103eff4b59/2026/08/28/ML-21185-8.jpeg" alt="Rajesh Babu Nuvvula" width="100" height="100"></p> +
</div> +
<h3 class="lb-h4">Rajesh Babu Nuvvula</h3> +
<p>Rajesh is a Senior Solutions Architect on the Worldwide Public Sector team at Amazon Web Services (AWS), collaborating with public sector partners and customers to design and scale well-architected solutions for cloud migration and application modernization. His areas of expertise include…+
</div> +
<div class="blog-author-box"> +
<div class="blog-author-image"> +
<p><img loading="lazy" class="alignleft size-full" src="https://d2908q01vomqb2.cloudfront.net/f1f836cb4ea6efb2a0b1b99f41ad8b103eff4b59/2026/08/28/ML-21185-9.png" alt="Ruchi Bhatia" width="100" height="100"></p> +
</div> +
<h3 class="lb-h4">Ruchi Bhatia</h3> +
<p>Ruchi is a Technical Product Marketing Manager at Amazon Web Services, where she drives product marketing for model training and customization on Amazon SageMaker AI and leads marketing for the AWS AI League. She holds a master’s degree in Information Systems Management from Carnegie Mell…+
</div> +
<div class="blog-author-box"> +
<div class="blog-author-image"> +
<p><img loading="lazy" class="alignleft size-full" src="https://d2908q01vomqb2.cloudfront.net/f1f836cb4ea6efb2a0b1b99f41ad8b103eff4b59/2026/08/28/ML-21185-10.jpeg" alt="Mark Ross" width="100" height="100"></p> +
</div> +
<h3 class="lb-h4">Mark Ross</h3> +
<p>Mark is the Chief Architect for AWS within Atos’ Cloud and Modern Infrastructure engineering function, bringing over two decades of technology experience across Financial Services, Government, Health, Utilities, and Media sectors. He is passionate about helping customers build, migrate, a…+
</div> +
</footer></content:encoded>+
+
+
+
</item>+
<item>+
<title>Tokenomics at scale: How Jamf built real-time spend enforcement for Amazon Bedrock</title>+
<link>https://aws.amazon.com/blogs/machine-learning/tokenomics-at-scale-how-jamf-built-real-time-spend-enforcement-for-amazon-bedrock/</link>+
+
<dc:creator><![CDATA[Arun Chandapillai]]></dc:creator>+
<pubDate>Tue, 01 Sep 2026 16:03:41 +0000</pubDate>+
<category><![CDATA[Advanced (300)]]></category>+
<category><![CDATA[Amazon Bedrock]]></category>+
<category><![CDATA[Customer Solutions]]></category>+
<guid isPermaLink="false">3e0afa841f894aea05b2ebf4260dd9e6fe0d4ab0</guid>+
+
<description>As generative AI adoption scales, cost governance becomes a top challenge. Learn how Jamf built real-time, per-user spend enforcement for Amazon Bedrock using IAM Customer Managed Policies, an Amazon Athena cost view, and a serverless AWS Lambda loop that applies tiered model limit…+
<content:encoded><p>Generative AI spend behaves unlike any cost line before it. Traditional compute scales with provisioned capacity. AI spend scales with <em>behavior</em>: a single engineer running an agentic coding loop against a premium model can burn more tokens in a…Diff display stops at 400 lines. The line counts above are from the whole diff. 48 lines shown here cut at 300 characters. The raw artifact at this commit is linked above.