Self-taught AI & automation engineer, after a decade of operating real businesses.
I design, build, and run production LLM and automation systems solo, end to end, on real servers. I ran businesses for years before I wrote a line of code, so I build for the outcome and can talk to the people who sign off.
Open to Forward Deployed, Applied AI, and founding-engineer roles.
In early 2025 I ran a limo company. I wasn't a developer.
Then the company website got hacked: 60,000 spam posts, tens of thousands of toxic backlinks, and no budget to pay anyone to fix it. I started deleting by hand and could see it would never finish. So I wrote code to do it.
That first script is the whole story. I didn't stop at the fix. Every problem after it went the same way: refuse the manual path, go programmatic, keep going until I understood the domain.
Fig. 1 · The cascade2025 → now
One hacked site, eight problems, each solved in code and then owned. Each step jumps to what it became.
02 Selected work
01 · live, distributed
Pocket fleet node
Production failures reach my pocket in seconds. A daily-driver phone turned into a first-class member of my server fleet: a carried edge node on a private Tailscale mesh, running an unrooted Linux node the servers can reach. Fail-closed by design: the phone holds no server credentials; the fleet initiates.
alerting
Self-hosted push wired into 30+ cron jobs
agentic
Voice to LLM parser to structured multi-tenant task filing; a notification bridge turns any app's alerts into triggers
measure
Drives the phone's real browser to run Lighthouse on real hardware over a real network
stack
Tailscale, Termux, Python, ntfy, adb, systemd/cron, FastAPI task API with a Claude Haiku parser
02 · production, four sites
Automated Content Engine (ACE)
A multi-site content platform that generates and maintains SEO content with no per-post human writing. I built it because the AI writing tools produced filler I wouldn't publish.
Eight safeguards, each added after a real incident.
pipeline
13 stages, 5-dimension quality scoring, 3-layer fact verification that catches hallucinations before publishing
isolation
Per-site database, credentials, and cron: zero shared state
dashboards
Flask + Alpine.js + ApexCharts, per-client auth, an embed mode for the public demo
Years of running ETON's Google Ads by hand, rebuilt against the API. Real-time account analysis paired with an LLM that recommends and remembers strategy, so recommendations build on prior decisions instead of starting cold.
backend
FastAPI, 70+ endpoints over the Google Ads API (read + write) and Search Console
A fully unattended daily pipeline: it picks a NASA open-data lead, writes vision-guided narration, checks that narration against the actual imagery with a second model pass, renders video, and uploads to YouTube. About $0.27 an episode.
In July 2026 I added two features, measured a decline, traced it to my own change, and reverted it. The code stays behind flags.
Cheap trips, not cheap flights. For 12 U.S. cities it prices the whole trip, airfare plus a week on the ground, and shows where your budget actually reaches: a week in Medellin costs less than a week in Las Vegas.
engine
Pairing, hub, budget, and comparison pages regenerated nightly from the database
stack
Python, Flask, SQLite, static pages served by nginx
A prompt-rewriting SaaS with accounts, token metering, and pricing tiers. It pivoted into a budget-first flight search engine, which became the cheap-trips product now live as DashAway.
stack
FastAPI, React, Postgres; later Flask, SQLite, YAML
07 · production, over 400 episodes
Nashville Restaurant Radio
A production pipeline that pulls new podcast episodes from RSS, transcribes them, writes styled show notes, and publishes static pages to the live site with no manual step.
watch
RSS feed, idempotent
publish
Episode page, homepage hero, episode index, and counter, all automatic
Customer-communication automation for local service businesses. Phase 1 shipped as an SMS product; phase 2, a voice-first AI receptionist, is built but not commercially live.
infraDigitalOcean Linux droplets on a private mesh
Self-hosted end to end: no platform between the code and the people using it.
03 More builds
Demo-site lead engine
Generated and deployed 1,058 sites across 538 local businesses: research, geo sweeps, payments, rep workflow.
distill-operator
A Slack bot to query the ACE platform in plain language. 34 automated tests, systemd deploy.
NASA Earth Pulse
Real-time NASA event data on a 3D globe. TypeScript, Next.js, Cesium, Fastify.
First-party analytics
Self-hosted web analytics (Python + SQLite) so client sites avoid third-party trackers.
Newsletter platform
A multi-tenant newsletter and blog engine with subscriber analytics and a spam filter.
NotebookLM batch generator
Headless batch production of audio, video, and slides from markdown decks, with retries.
Lead-research pipeline
Apollo API orchestration, rate-limit aware, to source corporate-travel prospects.
MealWise (in development)
Grocery price comparison: Kroger API plus a custom ALDI scraper (~95% accuracy).
Kennel data layer
SQL Server to SQLite ETL, ~643k rows, with validation that caught a 1,657-row truncation; 215 tables reverse-engineered.
Market-data recorder
A real-time recorder building its own tick-level history across ~555 instruments, with a full test suite.
Secure call-analysis pipeline
Recordings over a chroot-jailed SFTP drop, Whisper transcription, PII redaction before analysis.
04 Operator track
since 2006
Hillcrest Kennel
Partial owner, operations, family business
Hands-on operations in a family-owned Nashville boarding and grooming business since 2006.
since June 2016
Executive Transportation of Nashville
Owner-operator, luxury ground transportation
I own and run operations, fleet, staffing, and the P&L. I repositioned it from high volume to premium: revenue per service hour up about 2.6x on lower trip volume. It is also where I taught myself to build.
since 2025
Distill Works
Founder & Engineer
I built its production systems solo: ACE, the demo-site engine, and the operator tooling. I built them for my own business first, then turned them into something I run for clients.
05 Contact
I am a self-taught AI and automation engineer in Nashville. No CS degree; everything here I learned by building it and keeping it running.
Open to Forward Deployed, Applied AI, and founding-engineer roles.
