AI features that survive production
What this shop has shipped
Everything below is first-party work: built, shipped, and operated in production, not client case studies. Named client work appears here as clients approve it.
Catalog and ad automation at scale
- A Python pipeline that generates and QAs a catalog of more than 700,000 variants. It holds feed integrity across three marketplaces and four country accounts.
- 1,254 programmatically generated Amazon Sponsored Products campaigns, holding 17.9% ACoS lifetime.
- Eight years in production. Every integration had to be cheap to run and cheap to fix, which is why it still runs.
CRM and lifecycle automation
- A CRM and WordPress stack with behavior-triggered lead routing, webhooks, and retention flows.
- Lifecycle email sequences built, tested, and deployed in Make and Zapier, with the HTML hand-edited.
- Deliverability held steady through SPF/DKIM, list hygiene, and ongoing segmentation.
- Eventbrite registration integrated into the CRM across a recurring program calendar.
- Content pipelines on video APIs, cutting production cycles from days to minutes.
Where we sit in our own network
This is the general AI-development shop in a network of specialists. Ecommerce implementations live at ecommerceai.org, AI-automation career guidance lives at hub.aiautomation.engineer, and the Austin scene map is atxai.io. If your project is none of those specifics, it lands here: a product feature, an internal system, an integration.
First questions
What kind of AI development do you do?
We build custom AI features and systems for businesses, starting with LLM-powered product features: summarization, extraction, and assistants scoped to your data. That also covers internal automations with review gates, plus retrieval over your documents done properly. The fourth piece is integration work, wiring models into the systems you already run, including MCP-based tool integration. We do not do research-grade model training, and we do not ship anything customer-facing without human review gates and logging.
Do we need to be in Austin?
No — Austin is home base and the flag on the domain, not a service boundary. Engagements run remote-first, with the same review-gated process either way. Being Austin-local does buy you working sessions in person, and a team that shows up at the same meetups your engineers attend. Our sister site atxai.io maps that scene.
What does an engagement cost and how does it start?
We quote fixed-scope projects after a free scoping call: bring the workflow or feature idea and the systems it touches. You get back a written scope with deliverables, review gates, and handover terms. Or you get an honest "an off-the-shelf tool covers this; here is which." Production AI is mostly disciplined engineering, so we price it like engineering, not like magic.