From 1 Country to 9 in 60 Days — Scaling a Global AWS Community with AI and Serverless
How a non-developer TAM used Amazon Bedrock, Kiro, and serverless to build a 9-country community platform for $8/month.
By Ricardo Gulias | June 2026*
Two months ago, Golden Jackets Brazil was a single-page website celebrating 14 Brazilian professionals who earned all 12 active AWS certifications. Today, it's a global community spanning 9 countries, 6 active chapters, 89 members, and over 1,059 certifications — all built through AI-assisted development and running on a serverless architecture that costs less than $10/month.
I'm not a developer. I'm a TAM. Every line of infrastructure, every Lambda function, every CI/CD pipeline was built through conversation with AI — specifically Amazon Bedrock (via Kiro CLI) as my primary development partner. This is the story of how AI made it possible for a non-developer to build and scale a global platform.
The AI-First Approach
Let me be upfront: I didn't write this platform's code by hand. I designed the architecture, made every decision, debugged every production issue — but the implementation was done through AI pair programming.
My stack:
- Kiro CLI (powered by Amazon Bedrock / Claude) — primary development tool
- Amazon Bedrock — the AI backbone for code generation, debugging, and architecture decisions
- MCP Servers — custom Model Context Protocol servers that give the AI direct access to my AWS resources
- Steering files — persistent context that teaches the AI my project's conventions
This isn't "vibe coding." It's a deliberate methodology:
- I describe the problem (in plain language)
- AI proposes the architecture (I validate against Well-Architected)
- AI implements (Lambda, IAM policies, GitHub Actions workflows)
- I test in production (real users, real traffic)
- AI debugs (when things break at 2 AM)
The result: a TAM with no development background shipping production infrastructure at startup speed.
What Is the Golden Jacket?
The Golden Jacket is the highest recognition in the AWS certification ecosystem — awarded to professionals who hold all 12 active AWS certifications simultaneously. There are roughly 300-400 of these people worldwide.
Golden Jackets Brazil started as a way to connect them. What happened next surprised me.
The Numbers (Day 1 vs. Day 60)
| Metric | April 2026 | June 2026 |
|---|---|---|
| Members | 14 | 89 |
| Countries | 1 | 9 |
| Active chapters | 1 | 6 |
| Certifications tracked | 168 | 1,059+ |
| LinkedIn followers | 200 | 1,100+ |
| Podcast episodes | 0 | Published |
| Monthly AWS cost | ~$3 | ~$8 |
| Repos in GitHub org | 1 | 7 |
Countries: Brazil, Poland, UK, Chile, India, France + USA, Peru, Italy (onboarding).
How AI Built the Infrastructure
The MCP Server Pattern
I built custom MCP (Model Context Protocol) servers that give AI direct access to community operations:
1
2
3
4
5
6
7
# Golden Jackets MCP Server — Tools available to AI
tools = [
"list-members", # Query Cognito users by chapter
"chapter-status", # CloudFront/S3 health check all chapters
"invalidate-cache", # Clear CDN cache after deploys
"suggest-topic", # Submit article ideas via SNS
]This means I can say "list all members from the UK chapter" and the AI queries Cognito directly. No console, no CLI commands to remember. The AI operates the infrastructure through natural conversation.
Bedrock-Powered Automation
The AI doesn't just write code — it operates the community:
- Well-Architected assessments on client accounts using Bedrock-powered skills (3 critical findings in 60 seconds)
- Article generation — technical content drafted and refined through AI collaboration
- Infrastructure debugging — "the admin panel is showing Loading forever" → AI reads Lambda logs, identifies the merge conflict issue, implements auto-update-branch fix
- Chapter onboarding automation — a single shell script (
setup-chapter.sh) that provisions all AWS resources for a new country in 3 minutes, entirely AI-generated
The 72-Hour Platform Build
The entire platform — S3, CloudFront, Cognito, Lambda, API Gateway, DynamoDB, WAF, GitHub Actions CI/CD — was built in a 72-hour sprint through AI pair programming. No prior development experience required. Just architectural knowledge and the ability to describe what I needed.
The Architecture
Every chapter runs on the same serverless stack:
1
2
3
4
5
6
7
8
9
10
11
User applies via form
↓
API Gateway → Lambda (gj-apply)
↓
Creates GitHub PR (member card + photo)
↓
Chapter Leader approves in Admin Panel
↓
GitHub Actions deploys to S3 → CloudFront
↓
Member live on site in < 60 secondsPer-Chapter Infrastructure (AI-provisioned)
Each new country gets — provisioned by the AI-generated setup script:
- Route 53 hosted zone
- ACM certificate (wildcard SSL)
- S3 bucket with website hosting
- CloudFront distribution (PriceClass_100)
- DynamoDB table (visitor counter)
- Lambda function (counter + Function URL)
- Cognito group (chapter isolation)
- AWS Backup vault (daily, 7-day retention)
- GitHub Actions workflow (push → deploy → smoke test)
Total setup time per chapter: under 5 minutes.
Shared Infrastructure
- Cognito User Pool — single pool, isolated by groups
- Lambda gj-admin — admin API with chapter detection from Cognito token
- Lambda gj-apply — origin-based routing to correct GitHub repo
- API Gateway — single HTTP API, CORS per chapter domain
- SNS — notifications for applications, articles, sponsorships
- WAF v2 — rate limiting + bot protection across all distributions
AI as the 10x Multiplier
Here's what would have been impossible without AI:
| Task | Traditional approach | AI-assisted |
|---|---|---|
| New chapter infra | 2-3 days (manual) | 5 minutes (script) |
| Debug Lambda error | Hours reading logs | 2 minutes (AI reads CloudWatch) |
| Write GitHub Actions workflow | Half day + trial/error | 10 minutes |
| Implement MFA flow (all chapters) | Days of frontend work | 30 minutes |
| Create assessment PDF (client) | 4-6 hours manual | 15 minutes |
| Well-Architected review | Full day | 60 seconds with WA Skills |
The math: I estimate AI gave me a 5-8x productivity multiplier. Without it, this community would still be 1 country with 14 members and a static HTML page.
What Broke Along the Way
1. The PR Conflict Problem
When two members applied simultaneously, the second PR would conflict. AI-generated solution: after merging any PR, Lambda automatically rebuilds all remaining open PRs from the updated main branch.
2. The MFA Challenge
Cognito required MFA but the frontend didn't handle the challenge flow. AI diagnosed the issue from the error pattern and implemented the full MFA setup + verification flow across all 6 chapter sites in a single session.
3. Token Expiration (Silent Failure)
GitHub PATs expired silently, breaking admin operations across all chapters. AI identified the pattern, created a long-lived PAT, and updated secrets in all repos simultaneously.
4. The "Loading Forever" Bug
Chapter Leaders can't access GitHub directly. When a PR had a merge conflict, the Admin Panel hung. AI implemented auto-update-branch before merge — fix deployed via GitHub Actions in under 5 minutes.
The Chapter Leader Model
The key insight: I build the infrastructure with AI. Chapter Leaders run their community with zero technical skills required.
Each Chapter Leader gets:
- Write access to their GitHub repo
- Admin access to their Cognito group
- A step-by-step Operations Guide
- Full autonomy for day-to-day decisions
What they DON'T need:
- AWS Console access
- Programming skills
- AI tools
- My permission
Cost Breakdown
| Resource | Monthly Cost |
|---|---|
| S3 (6 chapter buckets) | $0.15 |
| CloudFront (6 distributions) | $0 (free tier) |
| Lambda (5 functions) | $0 (free tier) |
| DynamoDB (6 tables) | $0 (free tier) |
| Route 53 (6 hosted zones) | $3.00 |
| API Gateway | $0 (free tier) |
| Cognito | $0 (free tier) |
| WAF | $5.00 |
| AWS Backup | $0.50 |
| Total | ~$8.65/month |
A global community platform for 89 members across 9 countries. Less than a Netflix subscription.
The AI Development Workflow (for other TAMs/non-devs)
If you're a cloud professional who "doesn't code" but wants to build:
- Start with Kiro CLI — describe what you want, iterate on the output
- Use steering files — teach the AI your project conventions once, never repeat context
- Build MCP servers — give AI direct access to your AWS resources (read-only first)
- Trust but verify — AI writes the code, you validate the architecture against Well-Architected
- Ship fast, fix fast — production is the real test. AI helps you debug in minutes, not hours.
The barrier to building on AWS is no longer "can you code?" It's "can you design systems and describe what you need?"
What's Next
- AWS Summit São Paulo (September 2026) — first Golden Jackets in-person dinner
- 150 members by end of 2026
- 3 more countries onboarding (USA, Peru, Italy)
- AI-powered member matching — Bedrock to suggest connections between members with complementary skills
- Automated community insights — Bedrock analyzing engagement patterns across chapters
Lessons Learned
- AI doesn't replace architectural thinking. It replaces typing. You still need to know what to build and why. The 12 AWS certifications gave me the architectural foundation. AI gave me the implementation speed.
- Serverless + AI = solo founder superpower. Zero ops overhead + AI implementation speed means one person can build what used to require a team.
- Build for delegation from day one. If I'm the bottleneck, the community can't grow. Every system was designed so someone else can operate it without me — or without AI.
- MCP servers are the future of ops. Giving AI direct, controlled access to your infrastructure through MCP transforms how you operate. No more copy-pasting ARNs or reading CloudWatch manually.
- The community is the product, not the platform. The serverless architecture and AI tooling are enablers. The real value is 89 professionals across 9 countries helping each other grow.
The Stack
| Layer | Service | AI Role |
|---|---|---|
| Development | Kiro CLI + Bedrock | Primary implementation partner |
| Operations | MCP Servers | AI-operated infrastructure |
| DNS | Route 53 | AI-provisioned per chapter |
| CDN + SSL | CloudFront + ACM | AI-configured |
| Protection | WAF v2 | AI-tuned rules |
| Hosting | S3 | Zero servers |
| Auth | Cognito | Chapter isolation |
| API | API Gateway + Lambda | AI-written functions |
| Storage | DynamoDB | Counters, consent |
| CI/CD | GitHub Actions | AI-generated workflows |
| Notifications | SNS | Event-driven |
| Assessments | Bedrock + WA Skills | AI-powered reviews |
Golden Jackets is an independent community, not officially affiliated with Amazon Web Services. Built entirely with AI-assisted development on serverless AWS services.
Want to learn more? Visit goldenjacketsbrazil.com or find your country's chapter at global.goldenjacketsbrazil.com .
Enjoyed reading this content? Let the author know!
Your likes, comments, shares, and saves help creators reach more builders.
Loading recommendations
Loading article