Mission-critical commissioning is won long before integrated systems testing begins. When a vendor or the general contractor says "we did it this way on the last phase," I hold them to the specs, so equipment reaches functional testing, integrated systems testing, and turnover ready to pass.
Before commissioning, I was a Level 3 data center operations technician at a Tier IV hyperscale cloud provider, resolving the same tier of tickets as Level 4 technicians and restoring critical services under tight SLAs. That operating experience shapes how I read installation and system risk. At every data center site I've worked, everyone owns security: if you see something, you say something.
Rubicon Technical Services · hyperscale data center sites
Serve as the client-designated Commissioning Authority (CxA) and senior commissioning resource on site; train the Cx team and set its standards.
Own the Level 2 and Level 3 quality gates across the full power path, from medium-voltage switchgear to rack whips, before anything moves downstream.
Audit installations against design intent and specification: torque, grounding, lightning protection, FOD prevention, and readiness before energization.
Assess installed equipment and contractor test reports against the project specs, the owner's instructions, and the contract.
Write fact-based observations that cite the failed requirement, and track each one until it is closed.
Run Level 4 and Level 5 testing, including dynamic load bank testing, and verify 2N dual-path distribution and EPMS monitoring through turnover.
On my most recent project, ran Level 2 through Level 5 in parallel: my second L4/L5 deployment (under a lead, MV scope), while also serving as the only electrical CxA on an L2/L3 scope. Over a year of L2/L3 experience.
Data Center Operations Technician
Dec 2024 – Sep 2025
Tier IV hyperscale cloud provider
Advised and mentored Level 2 technicians coming up through an on-the-job training program, and served as their escalation point on complex hardware and network failures.
Resolved the same tier of tickets as Level 4 technicians.
Led Sev1 and Sev2 incident response to root cause, resolving about 95% of critical failures on the same shift.
Handled break/fix on about 400 servers a quarter.
Handled media destruction and chain of custody for drives and other hardware that required it. A clock started when I checked out a new drive into my name; any drive that entered the restricted area was treated as dirty, used or not, and drives I pulled went into my custody on the same clock. Every drive had to be installed or on its way to destruction or sanitization before the clock ran out, and unused drives were destroyed too.
Every destruction used two-person verification.
Kept the work logged for audit. Auditors checked the records and compared camera footage against the logs. Most of the auditing was automatic: workflows tracked every drive by serial number and tied it to the asset it was installed in, the security system knew where each drive came from, and people stepped in when an event came up.
Supported existing userscripts built by former employees.
Quality Control & Filtration Technician
Feb 2017 – Jun 2018
Crunch Pak · HACCP-controlled food production
Held stop-work and product-hold authority across the plant's inspection and filtration stations.
Ran food-safety chemistry and test-and-hold sampling with chain-of-custody records for lot traceability.
Earlier quality work: Quality Control Inspector, Mission Creek Carpentry (2005–2006).
Tools & AI
Built CxTools, a read-only browser userscript for Procore inspections and observations: it downloads reports and attachments, extracts inspections as structured data, lists unanswered items, and prefills observation forms for a person to submit. Two coworkers use it, one on a lite version and one using and testing the full version.
Use Claude every day and have built tools with it that produce consistent documents and spreadsheets.
My work NDA rules out public AI tools, so I use Amazon Bedrock in my own cloud account, where prompts and outputs aren't stored or used for training, with an AI data opt-out policy on the account. Every document first goes through a sanitizer I wrote. It strips client identifiers, removes unneeded photos and formatting, and converts documents to Markdown for AI use. It runs locally with no cloud uploads, or as a build and processing pipeline in the same account.
Built Claude skills that review commissioning inspections against the project specs and the owner's written procedures, and draft observations that cite the failed requirement.
Personal project: run my own cloud account as code (CloudFormation in Git, linted before each deploy), with a least-privilege identity for the AI agent that helps run it, a permissions boundary checked by an automated test, a multi-region audit trail, alerts on AI data opt-out changes, and an automatic budget brake.