I make the quality bar something the pipeline enforces, not something the team remembers.
Senior full-stack engineer — nine years of production web systems, five of them on automotive SaaS at national scale.
- TypeScript
- React
- Next.js
- Node.js
- Tailwind CSS
- Zod
- Vue.js
- PHP
- Python
- MySQL
- PostgreSQL
- MongoDB
- Redis
- WordPress
- AWS
- Vercel
- Docker
- Cloudflare
- GitHub Actions
- Jest
- Lighthouse
- Claude
- OpenAI
A quality bar moved out of the review conversation and into the pipeline, where it cannot be forgotten or argued with under deadline.
- Performance
- ≥ 95
- Accessibility
- ≥ 98
- SEO
- ≥ 95
- Best practices
- ≥ 90
- Largest contentful paint
- < 1500 ms
- Cumulative layout shift
- < 0.05
- Total blocking time
- < 150 ms
- Script weight
- < 260 kB
Any architecture claims reusability. The only honest test is building the second thing — and then reporting what actually had to change.
The platform behind dealer websites — built at Convertus, carried through the acquisition into AutoSync. Top all-time contributor across six and a half years.
Counted from the repository’s own history — private, so this is the one thing here you cannot click through and check.
Five years on the platform thousands of Canadian dealerships run their inventory on. The work I would point at first is the caching — and the outcome that mattered was stability, not speed.
- Custom in-house OOP PHP, not Laravel or Symfony
- Redis Object Cache Pro against AWS-hosted Redis
- No test runnerZero automated coverage in either project. The lead-capture path — the only revenue path — has no regression tests.
- No error monitoringOn total webhook failure a lead survives only as a log line, while the visitor still sees a thank-you page. Silent loss.
- No perf gate on the second siteRiflessi ships without the Lighthouse budget that protects DriftPilot, so regressions reach the live site undetected.