I build the tooling, guardrails and workflows that let engineering teams adopt AI inside real constraints, and I still ship code on the surfaces I know best, including native iOS. 13+ years as an engineer: Staff at BuzzFeed, Lead twice, and three regulated healthcare employers in a row (PHI separation, locked-down networks, AI budgets that run out mid-week). The through-line is the same everywhere. Make the system observable, then make it safe to move fast in.
How I work with Claude day to day →
Verification happens before the merge, on one change in isolation, so a failure names its cause and the fix is a revert. Where a prod-mirror tier exists for a real reason, I keep it and automate the promotions instead of deleting it.
Every tool I build makes something measurable first. I measured my own secret-masking plugin's false positives (17 of 26 code shapes) before letting anyone use it, then fixed the design.
MVVM at BuzzFeed, trunk-based branching and PR-time QA at Inspire, shared components at Surest. None of it came from a mandate. I built it, presented it, and the team picked it up.
Technical leadership means staying in the work: writing and reviewing code, setting standards, pairing, and crediting people's work out loud. I own outcomes end to end.
text-on-primary-background) directly in code, and replaced duplicated page code with shared components, so a style change is one edit.Cut preview database branches from 127 to 26 by hand, removing 84% of the database bill, then automated the cleanup with tests so it can't drift back.
Snapshot the database before any build that carries a migration, and fail the build if the snapshot fails. Opened, tested and merged in one afternoon.
Separate types for patient data and safe data, so logging and analytics can't carry PHI by accident. Governance built into the code rather than a review checklist.
Moved QA onto pull-request builds and automated CI on every PR. Releases shipped on their first release candidate.
Python/MySQL service with a Redis cache, an extensible type model, a legacy-shape endpoint for old app versions, and a 20-minute overnight cutover with no visible issues.
Parallelized iOS, Android and unit test workflows as independent pipelines, cutting build times 30%.
University of Minnesota, 2013
Most of my career was built on mobile, but most of my impact came from the systems around the app: delivery pipelines, backend services, documentation that regulators accept, and the way a team works with AI. At Devobsessed that's the job. I work on the platform and tooling side of AI adoption, and mobile is one of the surfaces I ship on.
I've spent most of the last decade inside regulated environments, so the governed enterprise trying to adopt AI against policy (not capability) is a seat I've sat in myself.
I do my best work on teams that give an engineer real ownership of an area, treat quality as part of delivery, reward trying things, and actually collaborate across roles. Regulated environments are a plus, not a hurdle.