In the development of Stojko Hub — the private operating system built for the entire family — that exact engineering discipline was put to the test. Unifying daily life into a coherent, multi-tenant digital hub, the platform integrates a personal nutrition and macro tracker, an intelligent meal prep and Sunday batch cooking engine, a Garmin health dashboard with AI-driven recovery analysis, and an automated GymBeam supplement scraper via MCP. Designed as a living system, future modules will orchestrate family hobbies and household chores scheduling — all running with zero monthly infrastructure waste while scaling seamlessly on demand.
$0.00
Monthly Infrastructure Cost
< 3.0s
Response Latency (Down from 60s+)
100%
PostgreSQL RLS Isolation
By splitting the Next.js 15 frontend on Vercel from Python FastAPI services on Google Cloud Run, the system scales to zero during idle hours. Migrating the Model Context Protocol (MCP) from legacy SSE to the modern stateless Streamable HTTP standard allowed microservices to sleep while guaranteeing instantaneous wakeups.
THE FILM ROOM // TACTICAL SCHEMAS
Sunday Batch Prep, Garmin AI Telemetry & GymBeam MCP
A 5x serving multiplier automatically scales shared family dinners while personal breakfast macros remain isolated at 1x. When cooking completes, an atomic PostgreSQL transaction transfers raw fridge inventory into freezer storage as individually portioned meals. Meanwhile, Python microservices on Google Cloud Run scrape live GymBeam supplement data via Model Context Protocol (MCP) and synthesize real-time Garmin recovery telemetry without cross-exposing private biometric records.
Next.js 15
FastAPI (Python)
PostgreSQL RLS
Google Cloud Run
GymBeam MCP
Garmin AI Telemetry
Hardening systems into a PCI-DSS compliant architecture: when an exposed service-role credential was discovered during an internal security audit, we ran a textbook incident response — rotated all production keys, audited 3 silently open tables, and rebuilt the authentication layer around JWT session tokens enforced strictly at the database layer.