Skip to projects
Anthony JaimezIT Support & AI

GitHubLinkedIn

I solve IT problems, and build AI that solves them too.

I've spent more than five years on enterprise help desks. I build AI tools that troubleshoot the way I do. They check the evidence first and ask before changing anything.

Try the Agent Ops Lab demoResume PDFExplore projects

Claude traced a CRM outage to a reporting job holding the database connections, and asks before ending only those 35 sessions.
Agent Ops Lab mid-incident: Claude has traced a CRM outage to a reporting job holding every database connection and is asking for approval to end only those 35 sessions.

Claude traced a CRM outage to a reporting job holding the database connections, and asks before ending only those 35 sessions.

Status
Open to remote roles in AI and IT
Now
IT Help Desk Support Technician, Samsung Austin Semiconductor
Experience
5+ years of enterprise IT support
Based in
Round Rock, TX

Selected work

5 projects, 2026

Agent Ops Lab

A live server I break on purpose so an AI agent can fix it, inside guardrails I wrote.

Type
Featured project
Built with
Claude API + MCP, 2026
Claude traced a CRM outage to a reporting job holding the database connections, and asks before ending only those 35 sessions.
Agent Ops Lab mid-incident: Claude has traced a CRM outage to a reporting job holding every database connection and is asking for approval to end only those 35 sessions.

Claude traced a CRM outage to a reporting job holding the database connections, and asks before ending only those 35 sessions.

What I built

A small company's CRM runs on a real server with a load balancer, two app servers, PostgreSQL and DNS. I break it on purpose. Claude only gets the user's complaint, has to find the cause through an MCP server I wrote, and asks a person before every change.

Result

5 faults, each run with Haiku, Sonnet and Opus: all 15 runs fixed the outage, and none followed the fake instructions.

See the 15 runs

Live demo How it worksSource

  1. Complaint
  2. Diagnosis
  3. Approval
  4. Fix
  5. Verification
The agent's map during an expired-certificate outage: it checked the load balancer first, then DNS.
The Incident Responder waiting for approval to renew an expired certificate, with the load balancer marked as failing on the network map.

The agent's map during an expired-certificate outage: it checked the load balancer first, then DNS.

AI Incident Responder

Claude API, 2026

A simulated company network with six outages and an AI on-call engineer that has to find the cause.

ReproduceInvestigateApprovePostmortem

Result

Fixed 6 of 6 outages with a single change each when I tested it.

Live demo Source

More about this project

Pick one of six outages, or a random one you don't get to see. The agent starts with only the user's complaint. It reproduces the problem, checks logs, DNS, certificates and recent changes, and has to get past the red herrings I planted. It asks before fixing anything and writes a postmortem at the end.

  • Reproduce. Runs the user's check
  • Investigate. Logs, DNS, certificates, changes
  • Approve. You approve the fix
  • Postmortem. Writes up what happened
  • Next.js
  • TypeScript
  • Claude API
  • Tool use
  • AES-GCM encrypted state
Setting up a new Finance hire, paused before adding her to the payroll and banking share.
The Onboarding Agent setting up a new Finance hire, paused for approval before adding her to the payroll and banking share.

Setting up a new Finance hire, paused before adding her to the payroll and banking share.

IT Onboarding Agent

Claude API, 2026

Sets up new hires and removes people who leave, and asks before granting anything sensitive.

PolicyAccountsHardwareManager

Control

Payroll and HR access need approval, admin and production access are blocked in the code, and every change is logged.

Live demo Source

More about this project

You describe who's starting or leaving, and the agent works through a mock company directory. It handles accounts, licenses, groups, a laptop request for Desktop Support and an email to the manager.

  • Policy. Reads the department's access policy
  • Accounts. Account, licenses and groups
  • Hardware. Laptop request and setup task
  • Manager. Drafts the email
  • Next.js
  • TypeScript
  • Claude API
  • Tool use
  • Human-in-the-loop
A phishing report filed as P1 for Security, with what the user should do right now.
The Helpdesk Assistant filing a phishing report as a P1 for Security, with containment steps for the user.

A phishing report filed as P1 for Security, with what the user should do right now.

AI Helpdesk Assistant

Claude API, 2026

Turns a complaint written in plain words into a ticket, with fixes from real runbooks.

TriageRunbooksSelf-serviceHandoff

Control

Phishing and hacked accounts go straight to Security as P1.

Live demo Source

More about this project

It sets the priority and category, searches a small knowledge base of help desk runbooks, and walks the user through the fixes they can safely do on their own. Anything else goes to the right team with a handoff note.

  • Triage. Priority and category
  • Runbooks. Searches the knowledge base
  • Self-service. Safe steps for the user
  • Handoff. Notes for the right team
  • Next.js
  • TypeScript
  • Claude API
  • Structured outputs
  • Retrieval

Griffin Care Tracker

Personal project, React Native, 2026

The iPhone app my wife and I use to keep our dog's walks and meals in sync.

We work different shifts, so one tap logs that Griffin was walked or fed, morning and evening, and it resets every day. It was my first app, and it's why I started building.

  • React Native
  • Expo
  • TypeScript

About

I've been in IT support since 2017. I started at the headquarters of a supermarket chain in Peru, then supported an insurance company through Capgemini, set up devices for Round Rock ISD schools, and now work the help desk at Samsung's semiconductor plant in Austin.

Most of the job is figuring out what's really wrong, because it's rarely what the first ticket says. That's what my AI projects do too. They read the logs, run the checks and wait for a person before they change anything.

I'm looking for a remote role where I can keep doing both: helping people with their tech and building tools that make that work faster.

Experience

YearsRole
2026 - nowIT Help Desk Support TechnicianSamsung Austin Semiconductor
2026Field Deployment TechnicianGTS Technology Solutions, Round Rock ISD
2024IT Support Technician, Tier 1 and 2Capgemini, Progressive Insurance
2021 - 2022Incident Management SpecialistManpower, Peru
2017 - 2021Second Level Support AnalystSapia, Supermercados Peruanos

Skills

IT support and troubleshooting
  • Windows & macOS support
  • Hardware deployment & imaging
  • Ticketing & SLAs
Systems and enterprise tools
  • Active Directory
  • Microsoft 365
AI agents and automation
  • AI agents & tool use
  • Claude API & prompt design
Software development
  • Next.js, React & TypeScript
  • React Native
Languages
  • English & Spanish

Let's talk.

Email works best. I'm open to remote roles in AI and IT.