Skip to content
View john-rocky's full-sized avatar

Block or report john-rocky

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
john-rocky/README.md

Daisuke Majima

Senior iOS · On-Device ML Engineer  ·  ex-Ultralytics  ·  🇯🇵 Japan  ·  Remote

Shipping LLMs and CV models that run fully on-device — iPhone · iPad · Apple Silicon Mac.
No server round-trip  ·  No per-token cost  ·  Privacy by default.

Open to roles   2-week sprint


📦 Featured

🧠 CoreML-Models — 1.7k★

The de-facto iOS Core ML model zoo. YOLO · SAM · LaMa · OCR · LLM · Stable Diffusion · depth · pose · style transfer.

🥽 RealityKit-Sampler — 176★

Runnable Apple RealityKit samples: anchors · physics · lighting · gestures · occlusion · mesh · audio.

🖼 SemanticImage — 156★

iOS image / video filter & ML utility — Vision + CoreML + AVFoundation, production-tested.

💬 CoreML-LLM — 130★

LLMs on Apple devices: CoreML · MLX · llama.cpp with quantization recipes.

🧽 lama-cleaner-iOS — 81★

iPhone / iPad on-device LaMa image-inpainting cleanup tool.

🎯 CoreML-YOLOv5 — 67★

YOLOv5 ported to iOS, with Apple Neural Engine notes.

+128 more public repos · browse all


🚢 What I do

Take AI models from "works in a notebook" to "ships as a real iOS app."

  • CoreML conversion + ANE optimization — PyTorch / HF / ONNX → CoreML / MLX / GGUF / ExecuTorch. FP16 · INT8 · 4-bit. Graph hand-tuning. KV-cache packing.
  • ARKit + LiDAR spatial pipelines — depth fusion · scene reconstruction · plane / object anchoring · RoomPlan · Vision overlay · Metal post-processing.
  • Production iOS apps — SwiftUI · UIKit · Combine · Swift Concurrency. 30 FPS LiDAR + on-device YOLO + on-device LLM. Sub-200ms recognition on iPhone 15 Pro.

🌐 Upstream contributions

swift-transformers executorch onnxruntime


📱 10 apps on the App Store

under my own developer account · apps.apple.com/developer/daisuke-majima

SnapMeasure · Models Zoo · super2x · Mask2Face · AnimateU · AIPainter · Blur. · Memosh · Tasmon · timestone


🛠 Stack

Swift SwiftUI Combine CoreML ARKit RealityKit visionOS Metal Vision MLX Python PyTorch Hugging Face ONNX TensorFlow C++ Kotlin Go TypeScript AWS Postgres


📊 GitHub


✉ Contact

📧 rockyshikoku@gmail.com  ·  🌐 john-rocky.github.io  ·  💼 LinkedIn  ·  𝕏 @JackdeS11  ·  ✍ Medium

Async-first · Plain text fine · Reply within a day


Currently solo-lead on production iOS apps combining ARKit + LiDAR + YOLOv8 + on-device OCR + on-device voice + multi-device backend sync at iPad scale. Detailed case studies under NDA.

Popular repositories Loading

  1. CoreML-Models CoreML-Models Public

    Converted CoreML Model Zoo.

    Swift 1.8k 166

  2. RealityKit-Sampler RealityKit-Sampler Public

    a sample collection of basic functions of Apple's AR framework for iOS.

    Swift 176 16

  3. SemanticImage SemanticImage Public

    A collection of easy-to-use image/video filter.

    Swift 156 27

  4. CoreML-LLM CoreML-LLM Public

    Run LLMs on Apple devices with CoreML, optimized for Apple Neural Engine + GPU

    Python 130 17

  5. lama-cleaner-iOS lama-cleaner-iOS Public

    iPhone/iPad version of lama cleaner

    Swift 81 20

  6. CoreML-YOLOv5 CoreML-YOLOv5 Public

    A sample project how to use YOLOv5 in iOS

    Swift 67 13