Date of Defense

4-24-2026

Date of Graduation

5-2026

Department

Statistics

First Advisor

Alvis Fong

Second Advisor

Wassnaa Al-Mawee

Third Advisor

Halil Ibrahim Dursunoglu

Abstract

Commercial cloud platforms rely heavily on subscription models, so everyday users fall into a "cloud trap" summarized by perpetual fees, forfeited privacy, and a lack of true asset ownership. At the same time, functioning consumer electronics get discarded as e-waste when they fail to meet the hardware requirements of modern operating systems. This project presents a sustainable and cost-effective alternative to both issues by repurposing a 2017 HP Pavilion laptop into a personal cloud and media server.

Despite severe hardware constraints, my software architecture rivals the utility of paid services. By replacing Windows with a headless installation of Ubuntu Server, system resources are preserved for essential tasks. To overcome the steep Linux command line learning curve, CasaOS is utilized as a lightweight dashboard to manage Docker containers. Containerization ensures applications operate in strict isolation, which prevents system crashes and hardware bottlenecks.

The deployed architecture completely replaces commercial services. Nextcloud handles file synchronization and backups, while Jellyfin and Navidrome manage movie and music streaming, respectively. The project also demonstrates the viability of running local Artificial Intelligence by combining Ollama and Open WebUI into a unified Docker container for a self-contained AI environment. Hardware performance is optimized using an "Engine and Trunk" storage methodology, where a fast solid-state drive manages application logic and a high-capacity mechanical drive handles bulk storage. Security and remote access are handled via Tailscale, a zero-trust Virtual Private Network that creates an encrypted tunnel for secure remote communication.

Ultimately, this implementation reinforces self-hosting as a highly accessible alternative to cloud storage subscriptions. The laptop chassis pulls under three watts of idle power, so the operational cost of the server is only $8 annually. Advancements in Large Language Models make troubleshooting seamless, so the barrier to entry has never been lower. This solution allows non-technical users to repurpose e-waste, save money, and reclaim absolute control over their files.

Access Setting

Honors Thesis-Open Access

Download

Included in

Data Science Commons

Share

COinS