Principal Software Engineer (Remote)

About Dalton Mills

Dalton Mills is the AI operating system for the trades. Any contractor can build the software their business actually runs on, in plain language, no code.

The platform you build on. Not the software you put up with.

Backed by Founder Collective, FPV, and angels including Gokul Rajaram.

Every operator in the trades has a list of things their software can't do. The features aren't hard. They're too specific for any vendor to build for one slice of one trade, so the list sits there for years. Every item got the same answer from their software company: too niche to build.

Ours is different. They build it themselves.

We give every trade the same agentic rails underneath. From there, each operator shapes the software around how they actually work. The same product, opened by a forty-truck plumbing operation and a two-person pest control shop, becomes two different products. Every field service tool treats them the same. We're the first platform built to learn the difference.

The role

A true 0 to 1. No legacy code, no inherited stack, no roadmap handed down.

You take a problem end to end. Design it, build it, ship it, watch it in production, iterate. Often in a day.

In one week you might ship customer-facing product, build the internal tool the company runs on, and make the architecture call everything sits on top of. The range is the job. Senior in substance: you set direction, not just close tickets.

The hardest problems here are unsolved on purpose. There's no playbook for what we're building, and the part nobody has figured out yet is the part you'll own. If that's the part you skip to, this is the role.

What you'll do

• Architect and ship across the full stack, from the platform itself to the agentic layer that does the work for the customer.
• Build agents as a first-class part of the product, not AI bolted onto software. You'll help define what agentic actually means in production.
• Build the internal tools we run on. When something is repetitive or fragile, your first move isn't a new SaaS subscription. It's to build it.
• Make the early architectural calls, the ones cheap to make now and expensive to unwind later, and set the bar for how we build.

What you'll bring

• A principal or staff-level track record shipping real software, full-stack. You've owned consequential systems end to end.
• Comfort moving fast, experimenting, and architecting under ambiguity, with the judgment to know when to think radically and when to ship incrementally.
• Strong fundamentals over any specific stack. We're early and flexible on tools, so we care how you think, not what's on your resume.
• You build with AI every day, or you're hungry to. Our whole thesis is that "technical" versus "non-technical" was a tooling problem. We live that internally.
• Low ego, high ownership. You see the problem, you own the problem, you move.

Helpful, not required: early-stage or 0→1 experience, agent or LLM systems in production, real exposure to the trades or vertical SaaS.

How we work

How we operate isn't separate from what we build. Four values, and the loop they form: Learn, Earn, Show, Go.

• Always be learning. The day you're too sure to be corrected is the day you start falling behind.
• Earn it daily. We earn trust by doing what we said and saying the hard part out loud.
• Show, don't tell. We get our opinions by building, not by meeting about building.
• Go make it happen. The problem you saw is now yours. No waiting for permission or a ticket.

We're building an agentic organization, not just an agentic product. We work in public: open channels, calls recorded by default, what one of us knows all of us know. Humans and agents work side by side. A company built this way moves at a speed a traditional one can't. That's an edge no competitor can copy.

Our stack

We're all builders here, whatever the title. You'll prototype with Claude or Codex before you pull in another engineer. Today we run on Claude Code, Codex, Linear, Vercel, Railway, Supabase, and GitHub. We're early and pragmatic. We'll use whatever gets us to the answer fastest.