FAA private pilot, in training.
Roughly 38 hours logged in a Cessna 172. Boulder Municipal as a home field, mountain training in the spring. Checkride scheduled for summer 2026.
~38h · BDU · 2025 —
About
I build product systems where technical judgment becomes operational leverage.
I’m the founder of Accent Labs, Inc., where I’m building Backlight: an independent measurement product for AI coding tools. Backlight helps engineering and finance leaders understand spend, adoption, output, quality, and license utilization without storing source code or adding instrumentation to engineering workflows.
The work is a continuation of problems I’ve spent years close to: systems that have to be fast, trusted, observable, and useful to people making decisions under constraint.
I started by building iPhone software before the App Store was the only distribution path: unofficial toolchains, jailbreak packages, and small apps released through Cydia. At JibJab, that early work became professional software work. I learned Ruby on Rails on JibJab.com, then became technical lead for JibJab Messages, the company’s mobile-first personalized GIF and sticker experience across iOS and Android.
At Apple, I worked first in iOS Platform Experience across Setup, Settings, iCloud sign-in, client frameworks, service boundaries, and SRP-based credential verification flows. In late 2018 I joined Vision Products Group, working on early visionOS System UI surfaces before managing and growing the team through product delivery.
That background shapes how I build now: fewer claims, better instruments, tighter feedback loops, and products that make technical work legible to the people responsible for funding and operating it.
How I work
Four habits worth writing down.
01
Hold the surface count small.
Every additional thing on screen is a thing you have to maintain, calibrate, and explain. Less is the constraint, not the goal.
02
Hire against the work, not the org chart.
Headcount that isn't anchored to a surface drifts. Surfaces that aren't owned by a person rot. Both are the same problem.
03
Instrument before you argue.
A leadership debate gets better when the system can produce a trustworthy read. The work is to make that read boring enough to use.
04
Write it down before you ship it.
The decision that can't survive a paragraph hasn't survived the meeting. Memos are cheaper than launches.
Practices
Two things I do around the work.
A KiCad bench in the basement.
Small two-layer boards. ESP32 e-paper desk object on Rev C. The work is slower than software and rewards patience in ways that translate.
Rev C · CERN-OHL-S · 2024 —