<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:dc="http://purl.org/dc/elements/1.1/">
  <channel>
    <title>DEV Community: Kieu Nguyen Dang Khoa</title>
    <description>The latest articles on DEV Community by Kieu Nguyen Dang Khoa (@kndkhoa).</description>
    <link>https://dev.to/kndkhoa</link>
    <image>
      <url>https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4208344%2F76fc2cb9-3358-432c-9189-c42098c1f12e.png</url>
      <title>DEV Community: Kieu Nguyen Dang Khoa</title>
      <link>https://dev.to/kndkhoa</link>
    </image>
    <atom:link rel="self" type="application/rss+xml" href="https://dev.to/feed/kndkhoa"/>
    <language>en</language>
    <item>
      <title>Building a RAG Pipeline for Enterprise: From Kiro IDE to Knowledge Base on AWS</title>
      <dc:creator>Kieu Nguyen Dang Khoa</dc:creator>
      <pubDate>Wed, 04 Nov 2026 08:24:56 +0000</pubDate>
      <link>https://dev.to/kndkhoa/building-a-rag-pipeline-for-enterprise-from-kiro-ide-to-knowledge-base-on-aws-5fi2</link>
      <guid>https://dev.to/kndkhoa/building-a-rag-pipeline-for-enterprise-from-kiro-ide-to-knowledge-base-on-aws-5fi2</guid>
      <description>&lt;p&gt;I have just completed a simple but effective RAG (Retrieval-Augmented Generation) architecture, integrated directly into the IDE to speed up internal knowledge search tasks.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fe9lc7oz65ux23u8gkmva.jpeg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fe9lc7oz65ux23u8gkmva.jpeg" alt=" " width="799" height="386"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Why is this architecture suitable for Enterprise?&lt;/p&gt;

&lt;p&gt;✅ Sensitive data doesn't leave the AWS Cloud — in line with the industry's stringent data control and security requirements. &lt;br&gt;
 ✅ Clear separation of Local Environment and AWS Cloud, easy to audit, easy to scale&lt;br&gt;
 ✅ MCP Server acts as a bridge to standardize, allowing many different "knowledge tools" (not just RAG) to be plugged in without having to modify the IDE&lt;br&gt;
 ✅ Practical applications: look up system architecture documents, business processes, compliance policies, operational runbooks — Engineers ask directly in the IDE instead of searching documents. &lt;/p&gt;

&lt;p&gt;This is a small piece of the journey of building an Engineering Intelligence Platform (EIP) that my team is implementing, where AI agents become real assistants for engineers, not just chatbots.&lt;/p&gt;

&lt;h1&gt;
  
  
  RAG #AWS #Bedrock #Banking #AIEngineering #KnowledgeManagement #FinTech
&lt;/h1&gt;

</description>
      <category>ai</category>
      <category>architecture</category>
      <category>aws</category>
      <category>rag</category>
    </item>
    <item>
      <title>From "Experiential Review" to Agentic Code Review</title>
      <dc:creator>Kieu Nguyen Dang Khoa</dc:creator>
      <pubDate>Wed, 04 Nov 2026 08:23:59 +0000</pubDate>
      <link>https://dev.to/kndkhoa/from-experiential-review-to-agentic-code-review-1ipo</link>
      <guid>https://dev.to/kndkhoa/from-experiential-review-to-agentic-code-review-1ipo</guid>
      <description>&lt;p&gt;One of the most common issues in my day-to-day Code Review work is that Code Review still relies heavily on TechLead's experience.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F5c4obtciumza7cy88tvn.jpeg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F5c4obtciumza7cy88tvn.jpeg" alt=" " width="800" height="286"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Reviewers must compare themselves:&lt;/p&gt;

&lt;ul&gt;
&lt;li&gt;Coding Convention 

&lt;ul&gt;
&lt;li&gt;Security Checklist &lt;/li&gt;
&lt;li&gt;API/SQL Standards &lt;/li&gt;
&lt;li&gt;Architecture Guidelines &lt;/li&gt;
&lt;li&gt;Best Practices &lt;/li&gt;
&lt;/ul&gt;
&lt;/li&gt;
&lt;/ul&gt;

&lt;p&gt;And as the number of PRs increases, the basic mistakes continue to repeat.&lt;/p&gt;

&lt;p&gt;Not to mention that each TechLead may have a different way of evaluating.&lt;/p&gt;

&lt;p&gt;One use case I'm testing is combining Amazon Bedrock AgentCore + Knowledge Base + GitLab + Kiro to build an AI Code Review Agent.&lt;/p&gt;

&lt;p&gt;The flow is as follows:&lt;/p&gt;

&lt;p&gt;Developer → Kiro → Agent → GitLab → Knowledge Base → Review&lt;/p&gt;

&lt;p&gt;The agent will:&lt;/p&gt;

&lt;ol&gt;
&lt;li&gt;Get PR/Diff from GitLab. 

&lt;ol&gt;
&lt;li&gt;Identify the relevant Engineering Standards. &lt;/li&gt;
&lt;li&gt;Retrieve Coding/Security/Architecture Rules from the Knowledge Base. &lt;/li&gt;
&lt;li&gt;Analyze code changes based on the rules themselves. &lt;/li&gt;
&lt;li&gt;Return the Finding, Severity, Recommendation, and Reference to the respective rule. &lt;/li&gt;
&lt;/ol&gt;
&lt;/li&gt;
&lt;/ol&gt;

&lt;p&gt;Thus, the Knowledge Base is no longer just a place to store documents.&lt;/p&gt;

&lt;p&gt;It becomes Engineering Knowledge that is fed directly into the Development Workflow.&lt;/p&gt;

&lt;p&gt;AI no longer only supports developers at every step, but begins to become part of the Software Development Lifecycle.&lt;/p&gt;

&lt;h1&gt;
  
  
  AI #AIAgent #AgenticAI #AmazonBedrock #AgentCore #SoftwareEngineering #CodeReview #DevOps #Kiro #AIFirst
&lt;/h1&gt;

</description>
      <category>agents</category>
      <category>ai</category>
      <category>automation</category>
      <category>aws</category>
    </item>
    <item>
      <title>AI Agent Debug</title>
      <dc:creator>Kieu Nguyen Dang Khoa</dc:creator>
      <pubDate>Wed, 04 Nov 2026 08:22:10 +0000</pubDate>
      <link>https://dev.to/kndkhoa/ai-agent-review-code-e39</link>
      <guid>https://dev.to/kndkhoa/ai-agent-review-code-e39</guid>
      <description>&lt;p&gt;Last week, a demo with the Dev team tested an AI Debugging Agent that was integrated directly into the developer workflow through the Kiro IDE.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F6xlxpcrqt2b6iri6becs.jpeg" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2F6xlxpcrqt2b6iri6becs.jpeg" alt=" " width="799" height="330"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Developers only need to include one question, for example:&lt;/p&gt;

&lt;p&gt;"Why did this API start failing after the latest deployment?"&lt;/p&gt;

&lt;p&gt;Instead of just giving an answer based on available knowledge, agents can orchestrate multiple steps to investigate:&lt;/p&gt;

&lt;p&gt;Kiro IDE → MCP → AgentCore Runtime → Debugging Agent → Skills → Internal Tools&lt;/p&gt;

&lt;p&gt;Some current capabilities:&lt;/p&gt;

&lt;p&gt;🔎 Retrieve API / application information&lt;br&gt;
 📂 Source code and GitLab testing&lt;br&gt;
 📋 Retrieve information from Jira/Confluence&lt;br&gt;
 ☁️ Audit AWS resources and CloudWatch&lt;br&gt;
 🧠 Maintain context via AgentCore Memory&lt;br&gt;
 🤖 Invoke LLM on Amazon Bedrock for Analytics&lt;br&gt;
 📊 Execution Observation via AgentCore Observability&lt;br&gt;
 🧪 Evaluate responses through AgentCore Evaluations&lt;/p&gt;

&lt;p&gt;The interesting point here is not just "AI answers dev eloper's questions".&lt;/p&gt;

&lt;p&gt;The agent began to be able touse tools, collect evidence from multiple systems, and perform a debugging workflow to get to the root of the problem.&lt;/p&gt;

&lt;p&gt;This is one of the ways tobring Agentic AI into the software engineneering, where AI not only supports developers to think, but alsobegins to participate directly inthe process of doing things.&lt;/p&gt;

&lt;h1&gt;
  
  
  AgenticAI #AIAgent #AWS #AmazonBedrock #AmazonBedrockAgentCore #Kiro #MCP #SoftwareEngineering #AIDevelopment #DeveloperExperience
&lt;/h1&gt;

</description>
      <category>agents</category>
      <category>ai</category>
      <category>debugging</category>
      <category>tools</category>
    </item>
    <item>
      <title>The app is PR Standard Reviewer</title>
      <dc:creator>Kieu Nguyen Dang Khoa</dc:creator>
      <pubDate>Mon, 26 Oct 2026 10:00:20 +0000</pubDate>
      <link>https://dev.to/kndkhoa/the-app-is-pr-standard-reviewer-4nll</link>
      <guid>https://dev.to/kndkhoa/the-app-is-pr-standard-reviewer-4nll</guid>
      <description>&lt;p&gt;Last Saturday I joined the AWS FCAJ meetup, and I got the chance to demo a project I have been building.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fbqbzs6msuro3zsbyh5x6.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fbqbzs6msuro3zsbyh5x6.png" alt=" " width="800" height="535"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;The app is PR Standard Reviewer, and it runs as an app on Kiro Crew. Kiro Crew is a platform for running AI agents and apps through a shared gateway, so instead of wiring up my own model and API keys, the app plugs into Kiro Crew and uses the model already available there. You install it, enable it, and it shows up right in the dashboard.&lt;/p&gt;

&lt;p&gt;What the app does: it reviews GitHub pull requests against your own coding standards, not a generic linter's rules. You maintain your standards as documents, point it at a PR, and it returns findings ordered by priority, each one citing the exact standard it came from with a suggested fix. The summary can be posted straight back to the PR.&lt;/p&gt;

&lt;p&gt;&lt;a href="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fliov8uf3ggabdzi5a6oe.png" class="article-body-image-wrapper"&gt;&lt;img src="https://media2.dev.to/dynamic/image/width=800%2Cheight=%2Cfit=scale-down%2Cgravity=auto%2Cformat=auto/https%3A%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Farticles%2Fliov8uf3ggabdzi5a6oe.png" alt=" " width="800" height="532"&gt;&lt;/a&gt;&lt;/p&gt;

&lt;p&gt;Demoing live is always a good forcing function. Explaining the design out loud made the core idea clearer: separate what to enforce (a Knowledge Base of standards) from how to review (Skills), and keep every finding grounded in a specific rule so reviewers can verify it. Building it as a Kiro Crew app also meant I could focus on the review logic instead of model plumbing.&lt;br&gt;
Thanks to the AWS FCAJ community and everyone who stopped by with questions and feedback.&lt;/p&gt;

&lt;p&gt;The project is open source under MIT (Link source on comment)&lt;/p&gt;

&lt;h1&gt;
  
  
  AWS #FCAJ #KiroCrew #CodeReview #DeveloperTools #OpenSource
&lt;/h1&gt;

</description>
      <category>agents</category>
      <category>ai</category>
      <category>automation</category>
      <category>github</category>
    </item>
  </channel>
</rss>
