Skip to content
brian@ijbm ~ /portfolio
$ ijbm boot --portfolio

Hi, I'm Brian Marcellino.

Computer Science graduate building AI, automation, and full-stack systems.

GitHub activity is loading. View GitHub profile

Navigation

/about/experience/projects/publications/contact
… /help · /infra · /skills
/themes to change vibe

Focus

AIGenAI/DL
BuildAuto/mobile
MarketsTrading/NLP
VisionCV research
> /infra

Production Infrastructure

Self-managed Linux production stack on a single Contabo VPS, fronted by nginx + Certbot, running multi-tenant ERP, multiple Next.js properties, dockerized services, and a full observability stack. Deploys use a release-symlink pattern with PM2 process management.

Host

Production VPS

Ubuntu 24.04 LTS · Linux 6.8 (PREEMPT_DYNAMIC)

  • edge · nginx + Certbot (Let's Encrypt) terminating TLS for every public domain
  • processes · PM2 — Node.js processes run as a non-root deploy user
  • containers · Docker + Docker Compose for stateful services and the observability stack
  • deploy · Release-symlink: ./releases/<timestamp-sha>/ ← current → atomic switch on success

Public domains · 3

rukost.id

Kost (boarding-house) marketplace

Docker ComposenginxPostgresRedisLibreTranslate
Open rukost.id →

Private surfaces · 5

Login-gated systems are described here, never linked from this public site.

Multi-tenant ERP — Metropolia tenant

Login-gated ERP (HR, finance, sales, attendance, tasks/kanban, IT business dev, civil, inventory, portal)

ReactTypeScriptNode.jsPostgreSQLPrismaSocket.IOPM2

Multi-tenant ERP — Midville tenant

Same codebase as Metropolia tenant; isolated database, isolated Node backend, distinct release pipeline

ReactTypeScriptNode.jsPostgreSQLPM2

Document management

Self-hosted Paperless-NGX archive backing ERP fund-request and attachment workflows

Paperless-NGXPostgresRedis

Workflow automation

Self-hosted n8n + n8n-MCP bridge for ERP integrations and scheduled jobs

n8nPostgresn8n-MCP

Observability dashboards

Grafana dashboards over Prometheus + Loki — internal access only

GrafanaPrometheusLokiAlloyBlackbox Exporter

Container services

rukost-webrukost-apirukost-nginxrukost-redisrukost-dbrukost-libretranslate
orqest-postgresorqest-paperless-webserverorqest-paperless-dborqest-paperless-broker
n8nn8n-postgresn8n-task-runnersn8n-mcp
erp-entry-monitoring-grafanaerp-entry-monitoring-prometheuserp-entry-monitoring-lokierp-entry-monitoring-alloy

Operating principles

  • ✦ Application-owned observability — instrumentation lives in the app repo; runtime stacks live in a sibling infra repo.
  • ✦ Strict /api/metrics cardinality — never label by user, email, lead, or ticket ID.
  • ✦ Login-gated surfaces are deliberately not linked from this public site.
  • ✦ Two-tenant ERP shares one codebase via release-symlink deploys with isolated Postgres databases.
  • ✦ Every public domain terminates TLS via nginx + Certbot; internal services proxy over the loopback interface.
→ /projects for the systems built on top of this · /skills for the stack