Serverless Booking Platform

Overview

A production-grade SaaS booking platform built entirely on AWS serverless infrastructure. The system handles appointment scheduling, staff management, customer communications, billing, and analytics — serving multiple businesses through a multi-tenant architecture.

The platform runs 106 Lambda functions across 20+ service modules, with a React frontend served via CloudFront, PostgreSQL for data persistence, and real-time notifications through email and WhatsApp.

Architecture

The platform is organized into five layers, each handling a distinct concern.

Entry Layer

Users access the platform through Amazon CloudFront, which serves the Vite/React SPA from an S3 bucket and routes API requests to API Gateway. Every API request passes through a JWT Authorizer Lambda that validates authentication tokens before reaching the core business logic.

The API Gateway is configured with throttling (100 req/s sustained, 200 burst), request tracing via X-Ray, and CORS policies scoped to the production domain.

Core (Private Subnet)

The heart of the platform — 106 Lambda functions written in Python, deployed inside a VPC private subnet for security. The functions are organized by domain:

• Auth: Login, register, JWT refresh, password reset, email verification
• Appointments: Create, cancel, update status, recurring bookings
• Availability: Set availability windows, holiday management, slot calculation
• Staff: CRUD, staff login, password management, scheduling
• Services: Service catalog, pricing, duration configuration
• Customers: Customer profiles, search, activity history
• Business: Business registration, settings, profile management
• Locations: Multi-location support, amenities
• Billing: Stripe checkout sessions, subscriptions, trial management, webhooks
• Calendar: Google Calendar connect, sync, disconnect
• Campaigns: Marketing campaigns, email blasts
• Forms: Custom intake forms, response collection
• Reviews: Customer review collection and display
• Waitlist: Queue management for fully booked slots
• Analytics: Booking trends, growth metrics, usage dashboards
• GDPR: Data export, account deletion, compliance
• Dashboard: KPIs, activity feed, health checks
• Admin: Super admin operations, feature flags
• System: Health checks, error monitoring

All Lambda functions access the database through VPC endpoints, keeping traffic off the public internet.

Data Layer

• RDS PostgreSQL — Primary database running in the VPC private subnet. Stores all business data, appointments, customer records, and configuration. Encrypted at rest with automated backups.
• S3 Photos — Object storage for business photos, profile images, and uploaded assets.
• SQS FIFO — Notification queue with exactly-once processing and a dead-letter queue for failed messages. Ensures no notification is sent twice or lost.

Notification Layer

Asynchronous customer communications processed through the SQS queue:

• SES (Simple Email Service) — Transactional emails: booking confirmations, reminders, cancellations, marketing campaigns
• WhatsApp — Direct messaging for appointment reminders and confirmations, with scheduled sends
• Notification Lambdas — Processor functions that read from SQS, format messages, and dispatch through the appropriate channel. Includes 2-hour and day-before reminder schedules.

Operations

• Secrets Manager — Stores JWT signing secrets, Stripe API keys, and database credentials. Rotated automatically.
• EventBridge — Scheduled triggers for recurring tasks: appointment reminders, trial expiry checks, KPI snapshots.
• CloudWatch — Centralized logging, custom metrics, and operational dashboards for all 106 Lambda functions.
• SNS Alarms — Alert notifications for error spikes, DLQ depth, and database health.

External Integrations

• Stripe — Full payment integration: checkout sessions, customer portal, subscription management, webhook processing for payment events.
• Google Calendar — Two-way sync: appointments created in the platform appear in the business owner's Google Calendar and vice versa.
• Apple Auth & Google Auth — Social login for frictionless customer onboarding.

Infrastructure as Code

The entire platform is defined in AWS CDK (TypeScript) across 7 stacks:

Stack	Resources
NetworkStack	VPC, subnets, security groups, VPC endpoints
DatabaseStack	RDS PostgreSQL, Secrets Manager
StorageStack	S3 buckets (frontend + photos)
ComputeStack	106 Lambda functions, layers, IAM roles
ApiStack	API Gateway, JWT authorizer, CloudFront
MessagingStack	SQS queues, SNS topics, SES
MonitoringStack	CloudWatch alarms, dashboards, EventBridge rules

Security

• VPC isolation — Lambdas and RDS run in private subnets with no public internet access
• VPC endpoints — S3, Secrets Manager, and SQS accessed through private endpoints
• JWT authentication — Every API request validated by a dedicated authorizer Lambda
• Secrets Manager — No hardcoded credentials; all secrets rotated and encrypted
• Encryption — Data encrypted at rest (RDS, S3, SQS) and in transit (TLS everywhere)
• GDPR compliance — Built-in data export and account deletion workflows

Key Design Decisions

106 single-purpose Lambdas over monolith: Each Lambda handles exactly one API operation. Cold starts are minimal (small package size), IAM permissions are scoped precisely, and deployments are independent — updating the appointment creation logic doesn't risk breaking billing.

FIFO SQS for notifications: Exactly-once processing prevents duplicate emails/WhatsApp messages. The dead-letter queue catches failures for manual review rather than silently dropping notifications.

Multi-tenant from day one: Database schema, API authorization, and data isolation are designed for multi-tenancy. Adding a new business is configuration, not code.

Results

• 106 Lambda functions across 20+ business domains
• 7 CDK stacks managing all infrastructure as code
• Multi-channel notifications — email + WhatsApp with scheduled reminders
• Stripe billing with subscription management and customer portal
• Google Calendar two-way sync
• GDPR compliant with data export and deletion built in
• Zero server management — fully serverless, scales automatically