Juyoung K.

Product & Mobile Engineer | AR / VR Expert | Computer Vision | Web

I build iOS, visionOS, web apps and computer vision systems. Based in San Francisco, California. COMPUTER VISION My most recent project turns Mixed Martial Arts fight footage into stats. Broadcast footage goes in, and a per-fighter analysis of every strike comes out. I've trained my own small models for each specialized purpose. It's a full pipeline: track the fighters, read their body positions, detect each strike, then score how confident the system is that it got it right. I've tested it on 10,000+ hand-labeled clips across several matches. It catches 90% of strikes. Random guessing at the same rate would catch 42%. For computer vision projects, it's particularly important to have proper test driven development and setting passing bars for tests before you begin them. iPHONE, iPAD, AND VISION PRO I have three apps on the App Store. One turns a single 2D photo into a 3D scene you can move around in (walk around in your memory on the Apple Vision Pro). It even opens straight in iMessage without having to download the app (App Clips). So I am well-versed with the smaller details of the Apple Platform as well. Another one is a live sports streaming app for the Apple Vision Pro, where you and a friend can watch sports together and actually sit next to each other in the virtual room using Spatial Personas. Apple's built-in video player can't do that, so I wrote my own and released the code for free. BACKEND Three years at Wells Fargo moving bank data to Google Cloud. I helped move hundreds of millions of financial records to modern platforms, nothing lost. I rewrote the processing layer and cut the run time by 60%. HOW I WORK I plan out the goals beforehand and openly tell you any caveats, unexpected, or bad results. I build systems that are honest instead of guessing. If you need a prototype, I'll build one you won't have to throw away when it works (handing off to other developers, etc). WHAT I'M GOOD FOR Anything that turns video or images into data. Training and honestly evaluating custom models. Designing how your data gets labeled. iPhone, iPad, and Vision Pro apps. And the cloud backend underneath all of it. Please feel free to reach out!