Sales systems
Finding the companies worth contacting, learning enough about each one that the message is worth reading, and getting it in front of the right person on a schedule. Wired into the CRM you already run.
- source
- enrich
- qualify
- sequence
- CRM
- 6 yearsSoftware engineering experienceFrom nuclear submarine simulators to neural nets in published research.
- 10×Less maintenanceRote build steps replaced with code generation across the engineering team’s internal tooling.
- ~$200kReclaimed per yearEngineering hours given back to the product instead of to the work surrounding it.
U.S. Navy simulation lab experience
NIWC Atlantic, 2019–2025LinkedIn (opens in a new tab)3D camera calibration with a custom neural net
K-Multiscope, MOCO ’19ACM (opens in a new tab)See some of my code yourself
dev-stack-bridge-toolGitHub (opens in a new tab)You’re handing this to an engineer.
Six years, one standard. I was a software engineer at Naval Information Warfare Center Atlantic, a U.S. Navy lab, and I was a direct employee there rather than a contractor. Software that has to keep running for years, under a team you eventually hand it to, is a different craft from software that works in a demo. It is the only kind I have ever shipped.
An engineer by training. Typed boundaries, tests, error paths, version control, and a written record of why each decision was made the way it was. None of that is an extra I bill for; it is how the work gets done when nobody is watching. My current code is public, architecture decisions log included, and you can read all of it before we ever speak.
And by instinct. I built my old team’s internal tooling because I could not leave the rote parts alone: rote steps replaced with code generation, complicated systems clarified with structure and clear documentation. Nobody assigned it. The figures above are what it returned. Bottail is that same instinct, pointed at your business instead of mine.
You work with me directly. I’m Kyle Stewart, and I am the whole company. The person who runs your audit is the person who writes the code and the person who answers on the Sunday. Nothing gets relayed, and nothing gets handed off after the scope is signed, which is why the scope I quote is the one I am personally on the hook for.
Three machines.
Described by mechanism, because that is what you are buying. No two businesses need the same one.
Content production systems
A pipeline from topic to published piece, with a review step a person actually controls. The machine does the fetching, drafting and formatting; the judgement stays where it belongs.
- research
- draft
- review
- publish
Custom integrations
For where the off-the-shelf tools stop and your process keeps going. Two systems that will not talk to each other, a format nobody supports, a workflow that currently exists in one person’s head.
- read
- transform
- reconcile
- write
Built with whichever the problem needs: n8n, Make, or custom code. The tool is a decision made after the problem is understood, not before.
Most automation works until the day it doesn’t.
The barrier to building this work collapsed a couple of years ago. What did not collapse is the difficulty of building something that survives contact with real data: a renamed column, a rate limit, an API that returns success and an empty body, a record with an apostrophe in it. When that happens on a Sunday someone owns the outage, and if the system was assembled rather than engineered, that someone is you, inside a tool you cannot read. It is the whole difference, and it is the part that is invisible in a demo, because every demo works.
What I will tell you on the call:
Not every process is worth automating. Some are cheaper to leave alone, and I will say which ones.
Anything that needs real judgement stays with a person. I build the machine around it, not over it.
Scope and cost are set after the audit, never before. I do not quote work I have not looked at.
- ValidationMalformed records are caught at the boundary and quarantined, not written downstream where you find them a month later.
- RetriesTransient failures back off and try again. Permanent ones stop and say which record, which field, and why.
- LoggingEvery run leaves a trace you can read without me sitting next to you. Silence means it is working, not that it stopped.
- HandoverBuilt in your accounts, on your infrastructure, documented. You are not renting access to your own process.
Audit first, then build.
01
Discovery audit
Thirty minutes. We map the process you want automated and find where it actually breaks, which is usually not where people expect.
02
Scoped project
A written scope before anything is built: what it will do, what it will not do, and what it costs. Nothing starts until you have agreed to it.
03
What follows
Some systems need someone watching them and some genuinely do not. If yours does, a retainer covers it. If it does not, I will tell you that.
Start with the audit.
One call and you will know whether there is anything here worth building. If there isn’t, I will say so.
Book a free audit (opens in a new tab)Thirty minutes, and no prep needed.