// interactive résumé · build 26.10 · Spartanburg, SC
I build with AI.
I read every line it writes.
I’m Grady Davis: founder of Syndicate, former night-shift GM, AI builder. Behind this text, a physics engine I wrote by hand is simulating hundreds of thousands of particles on your graphics card. No libraries. No frameworks. No templates.
move to push particles · click empty space for a shockwave · scroll to change the shape
01 / The floor
Before I wrote code, I ran the floor.
Nine years of retail, security and operations. Short-staffed shifts, inventory, lines out the door. It’s where I learned that systems break at the handoffs between people, which turns out to be exactly where AI agents break too.
- 2016Shift leadDollar General · team of 9
- 2018Assistant managerCircle K · team of 25
- 2021Security supervisorSecuritas · 50 officers, built the training program
- 2022Dealer relationsAmerican Credit Acceptance · network +10% in 12 months
- 2023General manager7-Eleven, Spartanburg · ran the store end to end
02 / The money
Then I learned how attention turns into revenue.
in TikTok Shop GMV, driven for brands including Intake Breathing and Rap Snacks.
In 2024–25 I led classes and coached smaller agencies, and was among the first to test new TikTok Shop features. Creator Ops Media, my TikTok Shop agency, is a consultancy now.
03 / The machine
So I built the whole back office. Then I built the company on it.
A CRM, client portal, Stripe store with subscriptions, quotes, booking, privacy-first analytics, uptime monitoring and a help center. PHP, SQLite and plain JavaScript, deployed on every push.
“AI does most of the typing. The engineering is mine.”
- Spec I write what it must do, and what it must never do.
- Review I read every diff. I catch the bugs the model is confident about.
- Research OWASP, NIST, the RFCs. Passkeys and Argon2id, written from the specs.
- Test Playwright suites, a mock Stripe server and fake mail before anything ships.
04 / The agents
Four AI agents run parts of the business. None of them run unsupervised.
The model proposes. Code enforces. A person approves.
- Outreach writer researches businesses and drafts a first email plus two follow-ups.
- Reply handler answers replies, or hands the conversation to me.
- Website builder drafts a paid client’s site for me to review before they see it.
- Blog drafter writes one draft a week. Nothing publishes without my OK.
05 / Local + cloud
Open-weight models, just as much as cloud agents.
Qwen runs on my Mac (16 GB) through Ollama and LM Studio. Kimi runs in the cloud. I use them for coding, agents and file management, and pick the model for the job: private files stay on the machine.
Scripted illustration of how I split the work, not a live model.
06 / Try to beat the code gate
Write a cold email. See if it gets past my rules.
Models are confident, and confident is not the same as right. So in my outreach engine the model only proposes; plain code decides what can be sent. This is a browser version of those checks. Fix the draft until every light is green.
Runs in your browser; nothing is sent anywhere. Modeled on my CRM’s outreach rules. Stuck? Links may go to syndicatemarketing.co, stusyn.com or their own site. And try sending to owner@donotcontact.example.
07 / Terminal
Poke around. Type help.
Try:
View this page’s physics shader (the real one, running right now)
08 / Let’s build
So. What should we build?
I’m looking for remote roles where AI meets real operations:
- AI solutions & implementation
- AI automation & agent ops
- Solutions engineering
- Forward-deployed engineering