Jeremy Badder
AI operations. I architect and direct AI agents to build and run business systems, and I check their work against reality. Spring Branch, Texas · open to remote · jeremybadder@gmail.com
jeremybadder.com · case studies, videos and a live work log
Summary
Twenty years in IT. I started in desktop support, looking after 200–300 users, many of them nurses. They trusted me enough that when a group of them moved to XL Health, they referred me, and I was hired there on their word. The last twelve years were production support for a healthcare contact-center platform, where being confidently wrong is expensive. Now I work AI-first: I direct Claude Code agents to build automations, internal tools and products, with human approval before anything important happens. I'm not a traditional developer. I decide what gets built, verify it, and own the result.
Experience
- Built Alfred, an assistant that drafts replies to the firm's client email for staff to approve and turns their corrections into rules. Case
- Connected it to QuickBooks: emailed bills become a one-click entry after a single-use approval.
- Run the firm's n8n workflows on a server I manage; changes are tested on staging against past emails before release.
- Built and run a subscription app for homeschool families: 16 paying families as of Oct 7, 2026. Case
- AI drafts support replies; I approve each one. Moved the back end to a new server with no duplicate customer emails.
Production support and platform engineering for an enterprise contact-center platform (Amazon Connect, Genesys, NICE CXone) in a HIPAA-regulated environment.
- Designed an AI workflow that pulls up runbooks and past incidents automatically, with human approval before any change, read-only access by default, and a mode that keeps patient data out of the AI's view.
- Wrote the team's standard for AI-assisted analysis: evidence kept separate from interpretation, every claim published with the query that produced it.
- Caught a 2x over-count in failure reporting, silent data truncation, and false alarms (a disk "91% full" that was actually 7%).
- Traced a production outage (1,111 errors in 15 minutes) to a specific change by correlating logs across four servers.
Hired on the referral of nurses I had supported at Health Dialog, who had moved here first. XL Health was later acquired by UnitedHealth Group.
Desktop and application support for a healthcare services company, many of whose users were nurses.
Tools I work with
- Directing AI
- Claude Code, Claude Agent SDK, MCP, context and prompt design, human-approval gates
- Automation
- n8n, Node.js, PowerShell, REST APIs, QuickBooks API, Stripe
- Platforms
- Cloudflare Workers and D1, Linux servers (systemd, cron), Tailscale, Astro
- Operations
- Splunk, AWS, Amazon Connect, Genesys, NICE CXone, ServiceNow, change management, runbooks (AI-assisted)