Containerized the services
Prepared the primary Flask backend and separate Face API for production Docker-based execution.
CASE STUDY 03 / PRODUCTION CLOUD DEPLOYMENT
Production deployment verifiedRubiconProduction delivery.
A production application deployed across a Vercel frontend and Terraform-managed AWS infrastructure, with containerized backend, face-processing, reverse-proxy and database services.
PRODUCTION TOPOLOGY● HEALTHY
PUBLIC FRONTENDVercel
React 19 + Vite
HTTPS
AWS EC2 / REVERSE PROXYCaddy
TLS termination + service routing
Docker network
BACKENDFlask
Application API
PRIVATE SERVICEFace API
InsightFace / buffalo_l
Data
PRODUCTION DATABASEPostgreSQL
Supabase integration
01 / OVERVIEW
Moving an application into production.
Rubicon became a practical deployment exercise in connecting application code, containerized services, infrastructure as code, HTTPS routing, database connectivity and production frontend hosting into one working system.
The deployment separates the public Vercel frontend from the AWS-hosted backend stack while keeping the face-processing service private behind the application infrastructure.
FRONTENDVercel
INFRASTRUCTUREAWS + Terraform
RUNTIMEDocker Compose
02 / MY ENGINEERING WORK
Infrastructure, containers, and production fixes.
Provisioned AWS with Terraform
Created the AWS deployment infrastructure and repeatable configuration required to host the service stack.
Configured Caddy HTTPS
Routed production traffic securely through Caddy to the appropriate backend services.
Connected production systems
Connected the Vercel frontend, production database, authentication configuration and AWS-hosted services.
03 / PRODUCTION VALIDATION
Verified at runtime.
RUBICON // PRODUCTION HEALTHSTATUS: OK
Flask backendHEALTHY
Caddy reverse proxyRUNNING
Face APIRUNNING
PostgreSQL serviceHEALTHY
Google OAuthENABLED
Face modelbuffalo_l / 512D
04 / PROBLEMS SOLVED
Production failures became engineering work.
Diagnosed and corrected production behavior around the detection endpoint so requests reached the expected processing path.
Resolved filesystem and container permission issues affecting the InsightFace model volume in the deployed environment.
Connected the frontend, reverse proxy, backend, private face-processing API and database through the production deployment topology.
Configured Caddy to expose the required application surface securely while keeping internal services behind the proxy.
05 / OUTCOME
A working production deployment, not just a local demo.
Rubicon demonstrates end-to-end deployment work across infrastructure as code, container orchestration with Docker Compose, HTTPS routing, database connectivity, application services, authentication configuration, health verification and production debugging.