We don't just describe the work. Here it is.

• A multi-tenant financial operations platform that ingests transaction data automatically, replacing manual entry
• Check printing directly from the system
• CPA-ready workbooks generated from live data, built to cut that $7,000 bill by thousands this year
• Role-based access, audit trails, and tenant-level data isolation, because financial data deserves nothing less

Hours of monthly manual work eliminated, records that stay reconciled all year, real cost reduction at tax time, and a treasurer who runs the whole operation without calling me. The first organization's build now serves additional organizations on the same codebase.

Client organizations' financial data never appears on a public page; that's a feature of how I work. So the screenshots below come from the Havenfield Soccer Coaches Association, a fully fictional demo tenant running on the same production codebase with the same isolation, seeded with three years of synthetic books. Every name, number, and bank is invented. The live demo is available on request, and it beats the description.

• Peer-to-peer recognition with points, a live activity feed, and GIF celebration
• Multi-recipient recognition, hashtag values, and a completion moment that makes giving feel good
• Multi-tenant architecture with configurable branding, one codebase, many companies
• Light and dark modes, mobile-first responsive design

Live in production with demo data, in pilot conversations with its first prospective customer. All names shown are demo accounts.

• A desktop app for Mac and Windows that reads 2,800+ company career pages, six remote-job sites, and the user's own job-alert emails, then removes the duplicates
• Claude reads the user's resume to suggest the job titles to search for, then scores every match from 1 to 10 with a one-line reason. Matches start appearing within seconds
• Spots "thank you for applying" and rejection emails and moves those jobs to Applied or Closed on its own
• Private by design: no account and no server. The search, the resume, and the notes stay on the user's computer

In early access with beta testers, free to download. Built with Claude Code and first released October 5, 2026, with four releases in the first two days. Every release is built and tested automatically on real Mac and Windows machines before it ships.

See Job Scout

• I set the direction. Every project starts with what the people using it actually need, written down before any code exists. Claude works from that, not from a guess.
• I test it like a user. I use what gets built, with realistic data on real devices, and try to break it. If something confuses me, it gets fixed before anyone else sees it.
• I check the claims. Totals, security rules, and what a tool says it does get verified, not taken on the AI's word.
• Lessons stick. When something goes wrong, the fix comes with a written rule or an automatic check, so the same mistake doesn't come back.

• The treasury. Bank statements are checked to the penny before a single row is saved, and no check is recorded until it's printed.
• Job Scout. Every release is tested automatically on real Mac and Windows machines. In its first week, testing caught a Windows crash and a Mac bug that would have broken every search, before either reached more users.
• Every screenshot. Client data never appears on a public page. What you see here is demo data.

You get AI speed without AI slop: a solo practice delivering on timelines that used to take a team, with a person accountable for every result. And it's teachable. Helping your team work this way is one of the things I do.

A short walkthrough video is coming to this page. Until then, I'll happily share my screen and build something in front of you.