NestPay Portal

Building an Internal Portal with AI-Assisted Development

NestPay Team — Nestlé

1

Agenda

🛠️ Part 1 — Methodology

  • Why we built the portal
  • Iterative prototyping with GitHub Copilot
  • Tech choices and approach

🏗️ Part 2 — Architecture

  • AWS architecture overview
  • Key components and data flow
  • Lessons learned
2

Part 1 — How We Built It

The Need: A unified internal portal for the NestPay team to monitor payments, manage configurations, inspect API definitions, and visualize metrics — all in one place.

🤖 Iterative Prototyping with Copilot

  • AI-assisted development using GitHub Copilot from day one
  • Rapid iteration: prompt → generate → review → refine
  • Copilot accelerated boilerplate: routing, CORS, cookie handling, AWS SDK calls
  • Human review for security-critical code (auth, sessions, secrets)

📦 Tech Stack

LayerChoice
FrontendVanilla HTML / CSS / JS
AuthMSAL.js + Azure AD (Entra ID)
i18nCustom locale system (EN, ES, FR, CA)
BackendNode.js Lambda — single function, custom router
ObservabilityNew Relic NerdGraph + QuickSight
3

Part 2 — AWS Architecture

                                 ┌──────────────┐
                                 │   S3 Bucket   │
                                 │  (Static UI)  │
                                 └──────┬───────┘
                                        │
┌──────────┐     ┌──────────────┐      │      ┌───────────────┐
│  Browser  │────▶│  CloudFront  │──────┤      │   Azure AD    │
│  (User)   │     │              │      │      │  (Entra ID)   │
└──────────┘     └──────────────┘      │      └───────┬───────┘
                        │               │              │
                        ▼               │              │
                 ┌──────────────┐      │      ┌───────▼───────┐
                 │ API Gateway   │      │      │   QuickSight   │
                 └──────┬───────┘      │      └───────────────┘
                        │               │
                        ▼               │
                 ┌──────────────┐      │      ┌───────────────┐
                 │   Lambda     │──────┼─────▶│   DynamoDB     │
                 │  (Node.js)   │      │      │  (Sessions)    │
                 └──────┬───────┘      │      └───────────────┘
                        │               │
           ┌────────────┼────────────┐  │
           ▼            ▼            ▼  │
    ┌────────────┐ ┌──────────┐ ┌──────▼──────┐
    │ Secrets    │ │ AppConfig│ │  New Relic   │
    │ Manager    │ │          │ │  NerdGraph   │
    └────────────┘ └──────────┘ └─────────────┘
                        │
                        ▼
                 ┌──────────────┐
                 │  Dispatcher  │
                 │  APIs (NestPay)│
                 └──────────────┘
4

Portal Capabilities

📋 Features

  • Auth: Azure AD SSO via OAuth 2.0, server-side sessions stored in DynamoDB
  • API Dispatcher: Transparent proxy to downstream NestPay microservices (dev / int / stg)
  • Config Viewer: Browse AppConfig applications, profiles & environments
  • Secrets Browser: List Secrets Manager entries
  • Multi-language: Custom i18n — EN, ES, FR, CA

📊 Observability

  • QuickSight: Embedded dashboard for payment analytics
  • New Relic Metrics: Accepted-payments rate queries with faceted filters
  • Log Search: Contextual log querying via NerdGraph API

🔒 Security

  • HttpOnly + Secure session cookies
  • CORS with origin validation
  • Secrets fetched from AWS Secrets Manager (never hardcoded)
5

Lessons Learned

✅ What Worked

  • Copilot drastically reduced boilerplate time (routing, AWS SDK, CORS)
  • Vanilla stack = zero build complexity, fast deploys (single zip)
  • Single Lambda keeps infra simple and cost-effective
  • CloudFront unifies frontend & backend under one domain (avoids cross-origin cookie issues)

⚠️ Watch Out For

  • Cross-origin cookies (SameSite, Secure flags)
  • Always review AI-generated security code
  • Lambda cold starts with 128 MB memory
6

What's Next? — Open Discussion

🚀 Go to Production?

  • The portal currently targets dev / int / stg environments
  • Should we promote it to production?
  • Prerequisites: security review, load testing, on-call ownership, PCI

🔐 Role-Based Access Control?

  • Today every authenticated user sees everything
  • Read-only vs. admin roles
  • Leverage Azure AD groups / app roles

💡 Other Features?

  • Config editing — push AppConfig changes from the portal
  • Alerting dashboard — surface New Relic alerts alongside metrics
  • Audit log — track who accessed what and when
  • Automated tests — integration + E2E coverage

🗣️ Your Input

  • What pain points does the team still have?
  • Which feature would bring the most value today?
7

Thank You

Questions & Discussion

8