Portfolio case study

Streamer

A fully functional Android and Cloud Server MVP built in a week with a Cursor AI agent. It already replaced Spotify in my own listening because it streams media from anywhere, with the kind of playlist, shuffle, upload, and background behavior users expect from industry apps.

Streamer — now playing

1/4

A week with Sawyer and Cursor

  • Deliverable product on day 1.
  • Expert knowledge combined with AI-paced development.
  • Better understanding of project infrastructure at scale.
  • Rapid testing.
  • Rapid product iteration.
  • A fully built MVP you use daily.

Project snapshot

Role: Product-minded builder (UX, system design, implementation)

Build mode: One-week MVP with Cursor AI agent

Primary goal: Ship something real enough to test demand and daily habit—not a mock

Focus areas: Rapid debugging, streaming UX, WebSockets + cloud architecture, OAuth

What shipped

  • Shipped a market-ready Android MVP in one week using a Cursor AI agent loop: tight feedback cycles on UX, architecture, and edge cases instead of weeks of boilerplate.
  • End-to-end personal media streaming from anywhere: uploads, playlists, shuffle, and playback that keeps running in the background without dropping sessions.
  • Industry-shaped stack for real-time and scale: WebSockets, cloud-backed services, and OAuth so the product is credible on day one—not a throwaway prototype.

Velocity, UX, and systems

Rapid debugging under real usage

Built an app I use every day. That forced fast root-cause fixes, clearer error handling, and playback paths that survive lock screens, task switching, and flaky networks.

UX as a first-class constraint

Playlist building, queueing, shuffle, and library browsing designed for one-handed use.

Production architecture

Built for the real world: realtime control, cloud coordination for sessions and media, auth that separates identity and current sessions from library data.

Why this kind of MVP matters

Compress time-to-learning

A week of focused building put a working product in front of real people. Features were road tested in hours or even seconds. All in the same time it would take a full team to build a single feature.

Real domain expertise

Infrastructure choices from human experience and product vision (database choices, cloud architecture choices, proper division of responsibilities between client and server) overcome the shortcomings of AI that keep non-technical vibe coders spinning in circles.