I supervise a remote team of 18+ agents and built the AI operations layer my team runs on: automated reporting, database-connected AI workflows, and data pipelines that turn hours of admin work into one command.
Engineer's brain, teacher's communication, builder's hands.
I'm Gerard, a licensed electrical engineer and former university professor from Quezon City, Philippines. Today I work as a shift supervisor for a remote chat operations company, where I manage a team of 18+ agents: performance reviews, coaching, onboarding, and daily reporting.
Here's the thing though. Most of that admin work isn't done by me anymore. I built an AI system that does it. Using Claude with live integrations into Notion, our company database, Chrome, and Excel, I turned raw notes and messy exports into finished, formatted, submitted deliverables that generate themselves on schedule.
I like understanding how a business actually works first, then building the smallest system that removes the repetitive 70% of it. When my integrations break, I debug and fix them myself. And because I spent years teaching and public speaking, I can explain all of it in plain language to anyone.
All of these run in real daily production at my current company. Not demos, not tutorials.
A scheduled AI workflow fires every day at 5:55 PM: queries the company database via REST API, filters to my team, summarizes feedback calls, coaching, and point systems, then publishes a formatted multi-section Notion report page.
15+ custom Claude skills that turn raw notes into finished documents: performance review recaps, coaching notes, task briefs, onboarding trackers. Each encodes company formats and SOP rules, so messy input still produces consistent output.
Merges two mismatched CSV leaderboard exports, computes revenue, response time, proactivity, and conversion per agent, compares against team medians, and auto-flags who needs coaching before problems grow.
Connected Claude to the company PostgreSQL database through its REST API. When queries failed with HTTP 300 foreign-key conflicts, I diagnosed the relationship ambiguity and fixed it with explicit FK hints, then documented the fix.
Notion-based onboarding tracker with competency checklists, completion percentages, and hard gates, plus a team wellness log. Both updated by AI from plain session notes with one command.
A self-contained interactive web page with five before/after case studies of these systems: the raw input, the one-command step, and the finished output, with sample data anonymized.
The same four steps, whatever the industry.
I sit with the process owner and map what decisions the work actually drives. Metrics and tools come after, not before.
Ship one workflow end to end, prove it on real data, then extend the pattern. No six-month big-bang builds.
pull → normalize → process → publishScheduled, zero-question automations. If a human has to remember to trigger it, it isn't finished.
Plain-language docs and training so the team can use it without me. Former professor, so this part I genuinely enjoy.
What I use daily and what I bring beyond the keyboard.
The short version.
Supervise 18+ agents. Built the AI reporting, coaching, analytics, and onboarding systems the shift runs on.
Executive assistant on sales and marketing, reporting directly to the CEO. Project supervision and coordination.
Technical lectures for managers, CEOs, and facility management professionals.
Taught engineering courses, class management, online and in-person tutoring.
Objection handling, sales strategy, client needs analysis, CRM lead management.
If you have a process that eats hours every week, there's a good chance I can turn it into one command or no commands at all. Fully remote, async-friendly, flexible hours.