---
title: "zenz-coreml"
description: "On-device multimodal Core ML experiment"
canonical: https://portfolio.skyline23.com/en/projects/zenz-coreml
language: en
content-signal: ai-train=no, search=yes, ai-input=yes
---

# zenz-coreml

> On-device multimodal Core ML experiment

## Overview

An on-device ML experiment that converts PyTorch models, including stateful KV-cache variants, to Core ML for memory-constrained iOS keyboard inference.

## Highlights

- Converted conventional and stateful PyTorch models with coremltools.
- Explored KV-cache models introduced for iOS 18.
- Targeted Apple CPU, GPU, and Neural Engine execution inside a keyboard extension.

## Metadata

- Type: Project
- Category: ML
- Years: 2024 - 2025
- Stack: Core ML, Swift, AI

## Links

- Source: [GitHub](https://github.com/Skyline-23/zenz-CoreML)

## Language versions

- [EN](https://portfolio.skyline23.com/en/projects/zenz-coreml/index.md) · [KO](https://portfolio.skyline23.com/ko/projects/zenz-coreml/index.md) · [JA](https://portfolio.skyline23.com/ja/projects/zenz-coreml/index.md)

- [Portfolio index](https://portfolio.skyline23.com/en/index.md)