Projects, experiments, and field work.

The Problem

Diagnosing Windows system health in field environments requires installing tools that may not be permitted, may not exist, or may themselves be compromised.

The Solution

Stateless PowerShell diagnostic framework. Captures baseline snapshots, compares drift, flags anomalies. Single-file. No install. No registry writes. No external dependencies. Runs on any Windows system.

The Problem

Downloading YouTube videos and other media had become tiring. I kept jumping from website to website just to find something that worked. As a cybersecurity student, I also had to pay attention to domain names, redirects, ads, and other risks on shady download sites. Some sites were full of ads, some looked unsafe, and some locked file formats or longer videos behind paywalls. At some point, the process stopped feeling worth the hassle. I wanted a safer, repeatable, terminal-based workflow I could control.

The Solution

I started building Mr. Roboto as an AI-assisted utility, inspired by Mr. Robot. The "robot" idea fit the goal: a small tool that could handle repetitive media-download work from the terminal. Mr. Roboto is a portable media downloader powered by yt-dlp and FFmpeg. Windows is the stable platform, supported through Batch and PowerShell launchers. Linux support is available in beta through a native Bash launcher, roboto.sh. The project focuses on making media downloads more repeatable, reducing reliance on random download websites, and giving users a clearer local workflow with logs, dependency checks, and platform-specific launchers.

The Problem

Cloud-hosted AI services raise privacy, cost, and dependency concerns. Exploring what is possible with local inference and retrieval on consumer hardware.

The Solution

Running local language models, building RAG pipelines, and experimenting with agent tool use. Focused on privacy-aware, offline-capable AI workflows.

The Problem

LinkedIn profiles are saturated with generic corporate language that obscures real skills and repels serious recruiters and collaborators.

The Solution

Open-source prompt framework that systematically rewrites LinkedIn sections — headline, about, experience — into high-signal, data-backed language. No code. No API. Distributed as a usable framework via GitHub.

The Problem

Most computer vision systems depend on cloud processing, creating latency, privacy, and connectivity issues. Exploring what is possible at the edge.

The Solution

Experimenting with Grove Vision AI V2 for on-device inference — object detection, classification, and visual monitoring without cloud dependency.