Seven hands-on projects across cybersecurity, IT infrastructure, and AI. Each one is written up with the environment, the steps I took, the evidence, and what I learned.
Detecting and triaging threats, and measuring exposure before an attacker does: security operations, SIEM integration, and vulnerability management.
A Python tool that pulls the latest alert from a live Wazuh SIEM, extracts MITRE ATT&CK context, and has OpenAI write an analyst-style triage report. First I had to bring the SIEM server itself back online.
Wazuh, Python, OpenAI API, MITRE ATT&CK
I scanned a Windows 11 VM in Azure twice with Tenable. Adding credentials took the results from 36 findings to 147 and surfaced 3 High-severity issues the network-only scan couldn’t see.
Tenable, Microsoft Azure, Windows 11, CVSS
Building and running the systems people depend on: Windows domains and Group Policy, Linux servers, networking and DNS, and virtualized environments.
A Windows domain built from two clean VMs: AD DS and DNS on Windows Server 2025, a domain-joined Windows 11 client, AGDLP group nesting, a permission-locked SMB share, and Group Policy for both the computer and the user.
Windows Server 2025, AD DS, Group Policy, DNS
A Rocky Linux server stopped answering ping, SSH, and HTTPS. I traced it layer by layer from the hypervisor to NetworkManager, found a valid profile that wasn’t active, fixed it with one command, and made the fix survive reboots.
Rocky Linux, nmcli, ip route, Xen Orchestra
An internet-facing multi-server platform I designed and ran from 2018 to 2021: XenServer and CentOS VMs, a BungeeCord proxy, port forwarding and a DMZ, custom Java plugins backed by SQL, and a team of 8 to 10.
XenServer, CentOS, Java, SQL
Making automation reliable enough to trust: AI-assisted engineering with real test coverage, and prompt design that keeps models grounded in verified data.
A real-time quantitative system I architected and built with AI coding tools: about 50,600 lines of Python, a 9,500-line React control room, and 335 automated checks that gate every deploy.
Python, React, SQLite, Claude Code
A multi-year study in making LLM output trustworthy, with sports analysis as the test bed: hallucinated rosters, controlled data sourcing, reasoning gates, and versioned prompts.
ChatGPT, Structured prompting
Want the one-page version? The resume summarizes all of this, including experience and certifications.
