Fifteen years of customer-facing technical support, then infrastructure. I've done technical support at a small Montreal pro-audio company since 2009, as senior technician from 2021 (supervising and training two juniors) and, since late 2025, as its only technician: 4,000+ resolved tickets, mostly Mac-based creative professionals with a driver, firmware, licensing, or network problem standing between them and a session. Outside work I run a 57-service self-hosted environment on Ubuntu, joined to a six-node Tailscale fleet, and I'm CompTIA Security+ certified with Network+ in progress. Alongside the support work, I spent nine years tour-managing and running front-of-house for Juno- and Polaris-nominated artists: 300+ shows in nine countries, console networks and RF coordination under pressure, no helpdesk. That's where the instinct to stay calm and diagnose comes from.

Cross-platform by necessity: macOS daily since 2005 and the platform I support professionally · Linux servers and a Linux laptop at home · Windows for PC builds and the occasional user bug report · iOS as a tailnet node.

How I build: the infrastructure and tooling repos below were built with Claude Code under my direction. I own the architecture and run them in production; each repo says so at the top. The bootcamp write-ups were prepared the same way from 2024 lab work.

• Shipping a fix for a stranger's Windows bug. The first issue on Subgeneratorr was a Windows user whose containers wouldn't start. Diagnosed CRLF line endings on the entrypoint from the log they pasted, shipped a .gitattributes + Dockerfile fix on a branch, and gave them copy-paste steps to verify before merging. Issue #1 → PR #2.
• Writing the runbook after the outage. A silently expired Tailscale node key took my home server off the tailnet. The postmortem became tailscale-fleet-watchdog: the incident, the two bugs the tests didn't catch, and why each design decision was made.
• Docs that explain the why. The networking doc for my stack walks DNS layers, address spaces, firewall rules and traffic paths the way I'd explain them to a user, not a machine.

A real, daily-driver environment, secured and automated end to end.

• Infrastructure Showcase: Reference architecture for the full 57-service stack: Docker Compose, nginx reverse proxy, Google OAuth2 on every web UI, wildcard SSL via DNS-01, UFW + fail2ban, a 5-wave tiered startup, and a split access model where public services go through OAuth2 and private ones bind only to the Tailscale address.
• Tailscale Fleet Watchdog: Per-node self-check plus a REST-API fleet audit for a six-node tailnet. Catches daemon logout, imminent or expired node keys, expiry-config drift, and offline always-on nodes. Bash, cron and launchd, 22 tests.
• Server Monitoring Suite: Container health, resource-threshold alerts, SSL-expiry checks, and log-health auditing, all with Discord alerting and cooldown/dedup logic.
• Restic Backup System: Modular, encrypted, versioned backups with offsite sync to Google Drive or Backblaze B2, retention pruning, and integrity verification.
• Boo Bot: Production Discord bot for the server's community. Rebrands every notification under one identity, suppresses floods, handles requests, and pulls trailers.
• Homelab Scripts: Tiered startup, inotify-driven file watchers, and a 6-source API fallback chain for metadata enrichment.

I direct AI agents to build software, and I evaluate those agents for capability and safety.

• AI Agent Evaluation at Mindrift. I author and calibrate adversarial and capability test cases for LLM-powered agents: indirect prompt-injection red-teaming on one side, MCP-driven capability evals gated by a CI/CD pipeline and an LLM-as-judge review on the other.
• Agent Dispatcher: Web dashboard for orchestrating headless AI coding agents (Claude, Gemini, Codex) against a queue of project tasks. I run it in production for my own projects.
• llm-eval-harness: LLM-as-judge harness that scores model outputs against a rubric and gates CI on the result. Runs fully offline with recorded fixtures.
• Subgeneratorr: Subtitle generator that orchestrates Deepgram Nova-3 or local Whisper with your choice of LLM for keyterm extraction and translation. Built to fill the gaps Bazarr can't cover.

• Penetration Testing: 3-day pen test covering SQL injection, RCE on Struts/Tomcat, Shellshock, and credential cracking.
• Splunk SIEM: Custom detection rules for brute-force, SQLi, and DoS, with incident-response writeups.
• Azure Cloud Security: Multi-VM deployment with NSGs, Jump Box architecture, and Ansible automation.
• IoT Vulnerability Analysis: TP-Link smart bulb security assessment, CVE reproduction, and MITM demonstration.

• CompTIA Security+ (SY0-701): earned 2026
• CompTIA Network+ (N10-009): in progress