BUILDS

Tools, sites, and systems I’ve built — many with AI in the loop. Same mindset as everything else here: figure out what you want to make, then make it reel.

AO Acquisition Platform

Screenshot to add: Campaign Studio — the KILL / SCALE / REFRESH decision cards.

A multi-tenant, white-label paid-acquisition platform for recruiting insurance producers — built and deployed as its own product.

White-label landing sites per agency, a Meta-integrated ad-decision engine, and an attributed lead pipeline, all behind one admin. The standout is a Campaign Studio that turns raw Meta metrics into plain-English calls per creative — KILL, SCALE, REFRESH, TEST NEW HOOK — with the reasoning and the next action. A master/variant landing CMS lets one master site cascade to every agency variant, and the lead pipeline tags every applicant by source and campaign with time-based PII masking and audit-logged reveals.

Built with Next.js on Vercel, the Meta Graph API with daily auto-sync, and Google OAuth admin access.

AO Forum Uploader

Screenshot to add: the in-browser pipeline mid-run — transcription with auto-suggested trim points.

An in-browser media pipeline that turns a multi-gigabyte Zoom recording into a trimmed, uploaded file — no desktop software.

Drop a recording and, entirely in the browser, it extracts the audio, transcribes it, auto-detects the call type and the trim points, lets you adjust, then trims and uploads straight to cloud storage. FFmpeg compiled to WebAssembly does the audio work client-side, paired with a Whisper speech-to-text model — then it reads the transcript to suggest the cuts, with a human confirming. Uploads use multipart parallel PUTs with presigned URLs and a graceful download-only fallback.

Built with FFmpeg-WASM, Whisper, and S3-compatible storage behind a JWT-authenticated uploader service.

AO Sales Leaderboards

Screenshot to add: the leaderboard demo — champion podium and a flip-to-reveal trading card.

A premium public leaderboard ranking the top 100 producers company-wide — embedded on the company site and used as a live recognition tool.

Admins drop a weekly spreadsheet and it publishes in seconds; producers get a tappable, shareable trading-card profile with a champion podium and tiered rankings. When it went company-wide I hardened it with caching so expensive reads run at most once every five minutes across all viewers, plus a middleware fast-path for anonymous embed traffic. Drop-and-publish updates, a referer-gated embed, and a safe demo with real faces and fake numbers keep it production-ready.

Built with Next.js and Postgres, with trading cards rendered on-canvas to a shareable image.

Sidewing

Screenshot to add: the clean video window beside the hidden controls window.

A native macOS video player for running training calls over Zoom, with a fully automated release pipeline.

When the host screenshares the Sidewing window, the audience sees only a clean video frame — the scrubber and controls live in a separate, hidden window. Multi-clip merge, trim, split, rotate, 0.25x to 4x speed, captions, picture-in-picture, and mp4 export. A GitHub Actions pipeline builds, packages, and ships a release in about three minutes with zero manual steps, and a signing-free self-updater solves the macOS App Translocation bug that breaks most do-it-yourself Mac auto-updaters.

Built with Swift and AppKit, with GitHub Actions for CI/CD.

AO Media Hub

Screenshot to add: the admin tool grid and the Google-gated login.

The platform everything else plugs into — a multi-user internal product for the whole media team.

Authenticated access, an admin area, and a couple dozen tools: calendar, training, invoices, leaderboards, team tracking, and a Meta Graph API social analytics dashboard. It runs real auth and access control — middleware route gating, Supabase auth, allowlisted Google OAuth, and guest versus admin views — and I ran and remediated a full security audit covering RLS hardening, CSRF protection, and dependency CVEs. The analytics dashboard syncs every two hours into a warehoused Postgres layer and reports engagement, reach, and follower trends.

Built with Next.js, Supabase (Postgres, auth, and storage), and Vercel.