React • TypeScript • Supabase • Stripe • PostgreSQL • Klaviyo • Vercel • Full-Stack Development • SaaS • Subscription Platforms • Web Apps

The outcome was a fully delivered MVP — a working subscription platform that the client launched and is now generating revenue from. A few things I'd call out as the key results:

A music sample subscription platform built to compete with Splice, WAVS, and Tracklib. The client owned a library of 10,000+ pre-produced audio samples and wanted to move from one-off transactions to a true subscription business — users subscribe monthly, receive credits, and spend credits to download samples.

I delivered the full MVP solo over 16 weeks for $20K: subscription billing, credit system, secure downloads, custom audio player, bulk admin tooling, and email automation. The client released full payment on milestone sign-off and re-engaged me for Phase 2 work immediately after launch.

I delivered the full MVP solo over 16 weeks for $20K: subscription billing, credit system, secure downloads, custom audio player, bulk admin tooling, and email automation. The client released full payment on milestone sign-off and re-engaged me for Phase 2 work immediately after launch.

The client had a working business selling individual audio files on a third-party storefront — but no recurring revenue, no platform of their own, and no way to scale beyond manual transactions.

The goal was to rebuild the business around a subscription model: monthly plans, credit-based downloads, and a content library that could grow to tens of thousands of samples.

Three constraints shaped every decision:

Solo execution — I'd own scoping, build, QA, and client communication

Non-technical client — the platform had to be usable end-to-end without developer support, including content management

I broke the project into 8 bi-weekly milestones, each with explicit acceptance criteria. The client signed off on every milestone before payment was released, which kept scope tight on both sides.

Core features delivered:

Three-tier subscription system (Starter / Pro / Elite) powered by Stripe recurring billing and the Customer Portal

Credit-based purchase system with variable per-sample pricing and idempotent re-downloads

Trial system — 5 free credits on signup to drive activation

Browse, search, and filter across the catalog (genre, BPM, key)

Custom audio player with interactive waveform visualization

Secure download delivery using signed URLs with time-limited access

Admin bulk upload — process 100+ samples at once via CSV + audio file matching, cutting content onboarding from hours to minutes

User dashboard with credit balance, subscription status, and download history (with no-charge re-downloads)

Klaviyo integration — 6 lifecycle events powering client-managed email automation

Phase 2A features — favorites, collections, and one-time credit purchases

A few decisions shaped the build:

Supabase over a custom backend. Auth, Postgres, file storage, and CDN in one platform meant faster delivery and ~$25/month hosting instead of $150+ on AWS. Row-level security policies kept user data isolated without extra middleware.

Stripe Customer Portal instead of custom billing UI. Saved 40+ hours of work, PCI-compliant out of the box, and matches the UX users already expect from Notion, Linear, and Vercel.

Klaviyo via webhook events, not direct integration. The platform emits 6 lifecycle events (signup, subscription, download, low credits, trial depleted, cancellation). The client owns email content and flows in Klaviyo's dashboard — no engineering needed to ship a new campaign.

Edge Functions for every external API call. Klaviyo and Stripe API keys never touch the frontend. All event tracking and webhook handling runs in Supabase Edge Functions with verified signatures.

Credit system idempotency.

Audio files are large. Downloads can fail mid-stream, get accidentally re-triggered, or be re-requested days later. I needed to guarantee that re-downloading a sample never re-charged the user — but also that the underlying file stayed securely gated.

I built a transaction ledger backed by a Postgres unique constraint on (user_id, sample_id, format). The first successful download writes a ledger row and deducts credits. Subsequent downloads of the same sample return a fresh signed URL but skip the deduction. Credit-affecting and file-access operations are wrapped in a single transaction so a partial failure can't desync the two.

Verified end-to-end during client QA: first download deducted credits and created a ledger row; re-download returned a new signed URL with no charge and no duplicate row.

Delivered on time and on budget — all 8 milestones hit on schedule

Full payment released on client sign-off

Business model transition complete — recurring subscription revenue replacing one-off sales.

"Worked with Sheng on the TSL subscription portal build across eight milestones. Code quality solid, deployments clean, work shipped on time without me having to chase him. Same-day replies on basically every thread, weekends included. When I flagged a tricky bug during QA, he diagnosed and shipped a fix fast, no back-and-forth. Pricing fair, scoping honest. Already lined up for the next phase. Easy yes for the next one."

— Jan