DEV Community

Cover image for I Built a CFP Copilot with Amazon Quick - From Prompt to Published App in Under 1 Hour
Yuuki Yamashita
Yuuki Yamashita

Posted on

I Built a CFP Copilot with Amazon Quick - From Prompt to Published App in Under 1 Hour

#amazonquick #cfp #i18n #ai

No code. Just a prompt. And it works.

The Problem

I'm an AWS Community Builder (AI Engineering) who speaks at conferences regularly. Every time a new CFP opens, I spend way too long copy-pasting old abstracts, rewriting them for new audiences, and trying to remember which projects would resonate with which crowd.

There had to be a better way.

Enter Amazon Quick

Amazon Quick (formerly Amazon Quick Suite, launched Oct 2025) is AWS's agentic AI workspace. Think of it as an AI teammate that can build web apps from natural language, run research, automate workflows, and connect to your data - all without writing a line of code.

The feature I used: Apps in Amazon Quick (currently in Preview).

What I Built

CFP Copilot - a web app that takes an event name, deadline, and optional theme keywords, then suggests your Top 3 talk proposals based on your past content. Multilingual (JP/EN/ZH).

How I Built It (The Whole Process)

Step 1: Sign up (free, 2 minutes)

Go to quick.aws.com → Continue with Google. No AWS account required. Free plan gives you 30-day access to Plus features including Desktop and Apps.

Step 2: Open Apps → Paste one prompt

Left sidebar → Apps (Preview) → paste this into the text area:

Build a web app called "CFP Copilot" that helps a conference speaker reuse 
and repurpose their past talk content.

The app should have:
- A text input for "Event name" and "Submission deadline"
- A text input for "Event theme / keywords" (optional)
- A button "Get CFP Ideas"
- An output section showing Top 3 talk suggestions

Each suggestion should include:
1. Talk title (in both Japanese and English)
2. Which past project or talk it connects to
3. A short CFP abstract draft (2-3 sentences)

The app should reference the user profile: yama3133, AWS Community Builder 
(AI Engineering), main themes: "AI agent autonomy vs human approval", "MCP", 
"Amazon Bedrock AgentCore", "Aurora DSQL".

Use a clean, dark-themed UI with AWS-inspired orange accent colors.
Enter fullscreen mode Exit fullscreen mode

The app was fully generated in about 2 minutes.

Step 3: Add multilingual support (one follow-up prompt)

In the app editor's chat panel:

Add a language selector at the top with options: Japanese (日本語), English, 
and Chinese (中文). Style it to match the existing dark theme.
Enter fullscreen mode Exit fullscreen mode

Done. Three languages, zero code.

Step 4: Connect your corpus for better suggestions

In Quick's chat, click + to attach a file → upload your personal knowledge base (I built mine with a Python script that collects GitHub READMEs, Qiita articles, and project memory files into one markdown doc).

The AI immediately started referencing specific projects, past talk themes, and even my blogging history.

Step 5: Publish

Top-right corner of the app editor → Publish. A URL is generated instantly.

⚠️ Note: Viewers need a free Amazon Quick account to access the app. Plan your article accordingly.

The Output (Real Example)

Input: re:Invent 2027, deadline 2027-05-01, theme: "AI agents autonomy"

#1 - エージェントに任せるか、人間が止めるか：本番AIエージェントの自律性設計パターン
Trust or Interrupt: Autonomy Design Patterns for Production AI Agents on AWS
Connects to: wallet-agent + Amazon Bedrock AgentCore

#2 - MCPで変わるエージェント設計：ツール・データ・承認フローを統一プロトコルで繋ぐ
MCP as the Nervous System of AI Agents
Connects to: Aegis (Slack approval plane) + MCP

#3 - グローバルAIエージェントの状態管理：Aurora DSQLで実現するサーバーレス分散エージェントメモリ
Stateful Agents at Global Scale: Serverless Distributed Agent Memory with Aurora DSQL
Connects to: DROPZERO + AgentCore

All three outputs were production-quality CFP drafts, not just vague suggestions.

Architecture

What Surprised Me

  1. The quality without corpus data was already good. Amazon Quick's base knowledge of AWS services meant it could generate relevant suggestions even before I connected my personal data.
  2. Multilingual output just worked. Chinese had 2 retries before succeeding, but eventually produced accurate output.
  3. The entire flow took under 1 hour from signup to published app.

Try It Yourself

  1. Sign up free at quick.aws.com
  2. Copy the prompt above into Apps
  3. Customize the speaker profile section with your own themes
  4. Upload your own content corpus for personalized suggestions

What's Next

  • Build a corpus collector script for your own GitHub/blog content
  • Connect the app to Quick Index for persistent memory
  • Wire up local MCP servers for deeper GitHub integration

Yuki Yamashita (@yama3133) - AWS Community Builder, AI Engineering 2026
Projects: wallet-agent | DROPZERO | Diagram AI

Top comments (0)