Dallin Theriault

Builds working software by directing AI development tools.

American Fork, Utah

Why I'm doing this

I architect and direct software systems rather than write the code by hand. I make the technical decisions, scope the work, drive AI coding tools to build it, and debug against real logs until it runs. I've shipped real, working projects this way.

I am not a software engineer, and I say so plainly. I come from field operations and client-facing sales, not a computer-science track. I taught myself to ship real systems by directing AI tools, and I run two businesses while I do it: Theriault Property Services LLC (DBA Sharpline Painting Co.) and Dallin Paul Ventures LLC (DBA Field AI).

I'm looking for a role where owning a problem end to end and communicating clearly matters more than a traditional coding pedigree.

Projects

I designed and directed each of these. The code was written with AI tools under my direction. Source and full READMEs are on GitHub.

Field

Multi-tenant SaaS with an AI voice receptionist

Field dashboard, Cascade demo tenant

A platform for service businesses: an AI agent answers calls, detects what the caller needs, books jobs, and captures leads, with each client's data fully isolated using PostgreSQL row-level security. I built it to a working demo and deliberately stopped at v0.6.3 after deciding the market was too saturated to chase, knowing when to stop is part of the point.

Built with VAPI · n8n · Supabase / PostgreSQL · Next.js · Twilio

Nillad

Local-first personal AI assistant

Nillad chat interface

A private assistant that runs entirely on my own hardware: tool-calling chat, reminders, two-way SMS, and a calendar, with data that never leaves the machine. Cut response time from over 45 seconds to about 1 to 2 by finding the real bottleneck and rebuilding the chat to talk to the local model directly.

Built with Ollama · Next.js · SQLite · n8n · Twilio · Tailscale

Field PM

After-hours emergency maintenance triage for property managers

Field PM app, Summit Ridge Residential demo tenant

An AI voice agent answers a property company's after-hours line, verifies the caller against residents on file, classifies whether it's a true emergency, and either escalates to the on-call technician with a tracked work order or defers it to business hours. Intake runs on serverless edge functions I verified directly against the live endpoint: authorized requests run the full triage chain, unauthenticated ones are rejected.

Built with Vapi · Supabase Edge Functions / PostgreSQL · Next.js

Lost Arcana. A first-person game I'm building on the side in Unreal Engine 5 and Blender, for fun and to keep learning new tools.