Gerardo Lazo
Arones

An end-to-end online booking platform for a premium transportation service. Built solo, from database schema and a rule-based pricing engine through payments, real-time route pricing, and automated SMS confirmations.

Architected the whole system from zero: Postgres schema, a rule-based fare + hourly pricing engine, a conflict-free scheduling model (per-vehicle timelines, buffers, atomic rebooking), Stripe payments, and SMS/email confirmations. Started with the boring docs (System Design, ERD, service-layer class diagram, MVP use cases) so the build itself stayed tightly scoped.

Configured a real engineering workflow on top, GitHub Org with branch protection, Vercel CI/CD, custom domain on Google Workspace, because shipping a one-person project should still look like a team made it.

An AI-curated live channel for a self-hosted Jellyfin library, scheduling an on-demand children's catalog into one continuous, automatically programmed stream.

Co-built a plugin that turns an on-demand Jellyfin library into a continuous, AI-curated live channel for children's content. I owned the backend: a Python/FastAPI service that the C# plugin talks to over REST.

The integration gap nobody had solved: Jellyfin couldn't natively serve the AI-scheduled stream. I engineered server-side video streaming straight from FastAPI to close it. All inference runs locally through Ollama, no external API in the loop.

A self-hosted infrastructure stack: a Proxmox host running 20+ containerized services, with networking, storage, and secure remote access managed end to end.

A hardlink-based media pipeline built around the *arr stack: Sonarr, Radarr, Lidarr, and Prowlarr feed qBittorrent (routed through Gluetun + ProtonVPN), while Bazarr, Shoko, and Cleanuparr handle subtitles, anime metadata, and orphan cleanup before Jellyfin and Plex serve the library.

It spans several Proxmox VMs: NetBird provides self-hosted mesh VPN, Nginx Proxy Manager and Traefik terminate TLS for internal and public services, and TrueNAS serves storage over NFS. A passed-through RTX 3090 runs local LLMs through Ollama — backing two self-hosted AI Discord assistants and n8n automations. Breaking and fixing GPU passthrough, NFS under load, and the occasional OOM'd VM has taught me more production Linux than any course.