Case study · ViralWorx.ai
Viral Worx
An AI growth workspace for creators on YouTube, TikTok, Instagram and Facebook: analytics, scheduling, bulk AI editing, media generation and coaching in one place.
- Role
- Founder / Full-Stack Developer
- Years
- 2025 — now
- Status
- Live

The problem
Growing creators juggle analytics, scheduling, content creation, channel health and monetization across separate tools, with no operations team behind them.
What it does
- Command Center: analytics across YouTube, TikTok, Instagram and Facebook in one view.
- A scheduler that publishes to all four platforms.
- Bulk Creator and Bulk Editor: AI-assisted titles, descriptions and tags across many videos at once.
- Media Creator: AI image and video generation through a separate renderer service.
- Coach Martocci: an AI coach for scripts, outlines and thumbnails.
- Public channel microsites generated from the database, plus machine-readable discovery files for AI agents (llms.txt, OpenAPI, MCP discovery).
- A companion mobile app, Viral Worx Studio.
How it's built
- Next.js and React, full stack, in TypeScript.
- Cloudflare Worker custom domains in front of a private Cloudflare Container that sleeps when idle.
- PostgreSQL through Prisma; background work on Google Cloud Tasks plus scheduled Worker crons.
- AI through Google Vertex AI (Gemini); Stripe billing with the webhook as the source of truth; Resend email.
Engineering decisions
- A "safe operations brain": automatic fixes come from a closed allowlist. Low-risk actions run on their own, medium-risk need approval, high-risk are only ever recommended, and every attempt, blocked or not, lands in a ledger.
- An append-only recommendation journal with typed evidence, including how fresh the underlying metrics are.
- Coach safeguards: plan gating, rate limits and prompt-injection checks.
- Media URLs pass a fetch guard that blocks internal addresses, so a post can't be used to reach private infrastructure.
- A release gate that compares the live build's commit with main, with rollback per layer.
- Cost discipline: one container that scales to zero, which the runbook estimates cuts idle hosting cost by about 90%.