How I Work.
The operating system behind how I build, ship, and run businesses.
Seven principles I actually operate by.
Not aspirational. These are how the work gets done day to day - at JointCommerce, at Causal IQ, at DoveMed, in the R&D lab.
Build, then run the business with it.
I don't separate building the platform from running the business. The same engineer who writes the code is the one explaining it to clients, defending margins, and shipping the next feature.
Automation is the operating layer.
At JointCommerce, automation wasn't an optimization - it was the operating system. 15 people managed 200+ accounts at 85% gross margins because the work didn't scale linearly with headcount.
AI fluency is a team standard.
At JointCommerce, daily AI-assisted development (Claude Code, Codex, Cursor, Copilot) was the team-wide operating standard. The team multiplier matters more than any one person's prompt skill.
First-party data is the moat.
Every advertising business eventually competes on data. The consumer app at JointCommerce existed because the advertiser side needed unique data to differentiate.
Two-sided platforms compound.
More consumer app users → richer CDP → better targeting → higher ROAS → more advertiser spend → reinvestment in the consumer app. The compounding loop is the business model.
Ship the simplest thing that works, then iterate.
The CTV measurement framework started as a single client's question. I shipped a Python script, found the 5-7 day signup acceleration, and scaled the methodology across the book.
Hand off code for continuity.
The CTV measurement codebase stayed at Causal IQ after I left. That continuity matters - work should outlive the worker.
How I structure a working day.
The shape changes depending on the season, but the rhythm holds: automation handles the predictable, I handle the judgment calls, and the building never stops.
Triage what the automation surfaced.
- Review automation outputs - pacing alerts, QA reports, performance summaries
- Triage what needs human judgment vs. what's already covered by the system
- Decide which engineering work moves the operating layer forward today
Focused engineering blocks.
- 2-3 hour blocks of focused engineering with Claude Code and Cursor
- AI-assisted pair programming for new features, refactors, debugging
- Codex and Copilot for cross-validation when the stakes are high
The human side of the business.
- Client calls, sales partnerships, team support
- Whatever needed human attention now that the morning was triaged
- This is where the platform meets the people who pay for it
Think long, read wide.
- Strategic thinking, write-ups, planning the next quarter
- AdTech industry reading - what's shifting in DSPs, identity, measurement
- Open-source code review - keeps the engineering instinct sharp
The tools I reach for every day.
Not aspirational. These are the tools currently open on my machine right now.
How I think about the hard decisions.
Three tradeoffs that show up constantly in operator-builder roles. Here's how I reason through each one.
At a 15-person company with $300K-$500K/month P&L and 85% margins, you can't afford slow OR sloppy. AI-assisted engineering closes the gap by raising the floor on speed without compromising quality. The Claude Code + Cursor + Codex loop means I ship faster AND with better test coverage than I did pre-AI.
I tend to build when (a) the off-the-shelf option is more expensive than my time plus opportunity cost, (b) we need ownership for customization, or (c) it becomes a defensible asset. CDP, custom pixel, CTV measurement framework all met that bar. Stripe, Salesforce, Tableau didn't.
Generalist as an operator; specialist when I'm deep in a problem. I can fix the Django bug in the morning and present to a client in the afternoon - the same brain context-switches because the operating layer is consistent.
Edges, honestly drawn.
Credibility comes from knowing where the line is. These are mine.
I'm not a research engineer. I don't build novel ML models from scratch. I integrate existing models into real products.
I'm not a designer. I rely on system fonts, Tailwind defaults, shadcn/ui, and good information architecture. The visual polish on this site comes from constraints, not Figma.
I'm not a tier-1 brand-name candidate. No FAANG, no MBA, no Fortune 500 logos - yet. The work has to speak for itself.
I'm not a manager-of-managers. The largest engineering team I've directly managed is 10 at DoveMed. Above that, I'd need a stepping-stone role.
What I'm exploring next.
Roles where the line between exec and IC is blurred - because that's where the operating layer gets built.
- Director / VP Ad Operations
- CTO / Founding Engineer (seed - Series B)
- Director Client Solutions / Sr. Solutions Consultant
- Builder-operator hybrid roles