A self-hosted home lab built to develop practical skills alongside my AI Software Engineering degree. It brings virtualization, home automation, media services, monitoring, and AI-assisted operations together on repurposed hardware.
Illustrated overview of the lab, not a dashboard screenshot.
01 / PURPOSE
I’m building a useful self-hosted environment while developing practical skills for my AI Software Engineering degree. The goal goes beyond installing applications: connect the services, automate repeatable work, investigate failures, and verify the results.
Repurposed hardware creates real constraints. Virtualization, resource limits, and careful storage handling help the lab support everyday services alongside experiments.
02 / SYSTEM DESIGN
A logical overview of the environment. Service placement is simplified rather than presented as an exact network topology.
03 / SELF-HOSTED SERVICES
AI-assisted operations
An assistant accessed through Telegram to inspect systems, troubleshoot problems, and develop automation. Hermes runs on the lab server and connects to an external AI provider; the language model is not hosted locally.
Home automation
Smart-home integrations and automation experiments, running in a dedicated virtual machine.
Media services
A self-hosted media library and playback service, with read-only access to NAS media.
Media services
Live TV stream and programme-guide management feeding Jellyfin.
Media automation
A DVD-processing workflow using MakeMKV and HandBrake, with output validation and verified transfers.
Networking
DNS filtering and local hostname resolution.
Observability
Service-availability monitoring alongside scheduled health checks.
04 / AUTOMATION
Python and shell scripts support service checks, media transfers, and application API calls. The DVD workflow makes correctness visible at each handoff.
Detect the inserted disc and identify its contents.
Process the disc with MakeMKV and HandBrake, within configured resource limits.
Check the completed video before treating the job as successful.
Copy files to the NAS and verify them with SHA-256 checksums.
Remove the local completed copy only after the transfer has been verified.
Trigger a Jellyfin library scan after a verified transfer.
Scheduled health checks send down/recovered notifications instead of constant status messages. Lightweight routine checks run without calling an AI model.
05 / FEATURED ENGINEERING PROBLEM
A DVD-processing job reported success, but the movie never appeared in Jellyfin. Working with my AI assistant, I followed the trail through application logs, database records, and the filesystem.
VALIDATION
The repair passed eight regression tests and a real short HandBrake encoding test inside the ARM environment. Subsequent DVD jobs completed through to NAS storage and Jellyfin.
06 / OPERATING SAFEGUARDS
Key-based SSH for server administration; credentials kept out of source code and public documentation.
Separate virtual machines and containers for different services.
CPU, memory, and process limits keep DVD processing from overwhelming the server.
Jellyfin reads NAS media through read-only access.
Scheduled Proxmox backups are stored on a separate physical NAS.
Sensitive AI-assisted actions require approval before execution.
These are practical safeguards. They are not a claim that the lab is fully hardened or that the backups have been disaster-recovery tested.
07 / REFLECTION
A success report is not enough. The expected output must exist, be valid, and reach its destination.
A failed move must stop cleanup, or an error can become lost data. Verification belongs before deletion.
AI can help investigate and build, while lightweight routine checks and human approval keep operations deliberate.
The lab remains an ongoing environment for building automation, investigating failures, and checking that the result matches the intended outcome. Backup restoration still needs hands-on validation.
↗ / KEEP EXPLORING
ACTUAL GAME INTERFACE / STOCK MARKETMaintaining and extending a persistent multiplayer RPG: stateful gameplay, transaction-backed systems, controlled releases, and scheduled world updates.
LET’S BUILD SOMETHING USEFUL
Whether you're hiring an engineer or exploring a custom software solution, I'd love to hear what you're working on.