Mumu Li (n3xta)

RewindpixSince summer 2025

$1.5M: raised on Kickstarter
~12K: accounts, still growing
8,600+: monthly active users

A screen-free digital “film” camera and its companion app. I built most of the software myself and own it end to end: the iOS and Android app, the link to the camera and the backend. I also hold the design patent, and I’m responsible for the design system and the website, our landing page and community hub.

Three friends laughing on a sunny street, two holding Rewindpix cameras — rewindpix.com

Role

Co-founder, CTO

Did

App: UI/UX design, software engineering, embedded, database, revenue, CI/CD. Basically everything.
Website: Shopify, UI/UX design, full stack, community platform

Stack

Design: Figma and Code Connect, Flutter golden tests, Rive, colour science and LUTs, accessibility
Engineering: Dart and Flutter, GLSL fragment shaders, C++, Swift, Kotlin, HTTP/XML, image processing, Supabase
Product: Xcode, Gradle, Apple and OAuth sign-in, RevenueCat, universal and app links, QR scanning, code signing, App Store and Google Play releases, i18n, GitHub Actions, golden and integration tests

Links

Press

Case studies

UI/UX redesign ↗
System designLocked, recruiters only
Graphics processing pipelineLocked, recruiters only

The Rewindpix camera in cream, front view — The camera, in two colourways

The Rewindpix camera in navy, front view — The camera, in two colourways

Annotated line sketch of the camera, its controls labelled by hand — Product design draft: hot shoe, shutter, winder, viewfinder and a three-way film-filter slider

A roll developing in the app’s darkroom

Developing, frame by frame

New stocks arrive sealed: Cyber Warm 800, PT400 and Redscale

A sample frame shot on the CYBERWARM film stock

The film shelf

Dozens of Rewindpix app screens tilted in a grid: rolls, film stocks, the darkroom — The app, screen by screen

The community hub on rewindpix.com: a look passed from one member to the next

A wall of member photos in perspective, each tagged with its film stock

Search results for film photos on the desktop site, light theme

A film strip with sprocket holes and six frames from sample rolls — Frames from sample rolls

pixelatdSince spring 2026

262: ITP and IMA courses, synced every week

Course reviews for NYU ITP and IMA, run as an internal community site. Students rate workload, see what a class actually asks for and plan their semester before registration. I built it alone and run all of it: the design system, the web app, the pipeline that builds the catalog from Albert, the bulletin and the ITP forums, and the deploys.

Role

Founder, designer, engineer

Did

UI/UX design, design system, brand and type, full stack, database and access rules, data pipeline, i18n, CI/CD, promo graphics

Stack

Design: Figma, with a generative plugin I wrote for the pixel grid. Geist and Geist Pixel, custom rasterized display glyphs, CJK pixel-font subsetting, data-driven covers, accessibility
Engineering: Next.js 16, React 19, TypeScript, Tailwind 4, Supabase (Postgres, RLS, RPC), Python scrapers and merge logic, Claude API translation with a Postgres cache
Product: Cloudflare Workers and OpenNext, GitHub Actions for the weekly catalog sync and deploys, Supabase Auth, Resend, cron-driven points and badges, English, Chinese and Spanish dictionaries translated by hand first

Links

pixelatd.io ↗

Case studies

UI/UX redesign
Catalog pipelineLocked, recruiters only

Poster: a wall of pixelatd review cards, course search and radar charts, with classmates’ pointers over them — Data-driven course covers, and the poster

Six aspects on a radar, and where a course sits between seminar and production, chill and grind

A course profile

Course barcode

Websites I built

Processing Community Day 2026

miracool 2026

HackNYU 2026

NYC poop map

Artworks

They sit at the bottom of the page, but these are the works I’m proudest of. I’m learning to call myself an artist.

Your Voice in Landscapeyourvoice.landSpring 2026

16.5K: voices mapped, from the US, China, Australia, Japan, the UK, Spain and further

A machine-learning web app that listens to you speak for a few seconds and finds your voice a place on Earth. It measures six things about your voice: range, tessitura, timbre, colour, weight and volume. Each one becomes a geological image, and together they settle on one real coordinate. You get a name for the place, a short description of its terrain and a square map card to save and share.

Role

Designer and engineer

Did

UI/UX, visual identity and typography, the React front end, Mapbox integration, the acoustic feature pipeline, structured LLM placement, the per-profile result cache, the test suite, CI/CD and VPS deployment

Stack

Design: Figma, Mapbox GL
Audio and DSP: Python, Praat (Parselmouth), librosa, pydub and ffmpeg, NumPy and SciPy. Pitch tracking with percentile range, F1 formant analysis, harmonics-to-noise ratio, spectral bandwidth, RMS dynamics and semitone quantization
Engineering: React, TypeScript, Vite, Flask, gunicorn, SQLite