Mobile On-Device AI: Swift/Core ML and Kotlin/Android
For iOS and Android engineers who need to ship AI features that run on the phone, decide when to fall back to the cloud, and keep apps fast, small and private.
The curriculum is being prepared and reviewed. Enrollment is not open yet.
About this course
Phones now ship with system language models and capable neural engines, which changes what a mobile team can build without a server round trip. It also adds new failure modes: features that only work on some devices, models that are still downloading, tight context windows, battery and thermal limits, and app-size budgets.
This course covers both platforms in parallel. On iOS you use the Foundation Models framework (guided generation with @Generable types, tool calling, availability checks) and Core ML for your own models, with Core ML Tools for conversion and MLX Swift for experimentation. On Android you use the ML Kit GenAI APIs on top of AICore and Gemini Nano, and LiteRT and LiteRT-LM for custom models. You compare cross-platform options (ExecuTorch, ONNX Runtime Mobile), learn how to quantize and profile models on real devices, design hybrid flows that fall back to a cloud model through Firebase AI Logic or your own backend, and handle privacy, consent and store review. Platform features are stated as of October 2026; capability checks are taught as a habit precisely because availability keeps changing.
What you will be able to do
Decide whether a feature should run on device, in the cloud or as a hybrid, using latency, privacy, cost and device-coverage criteria
Implement a structured-output feature with Apple's Foundation Models framework, including availability checks and graceful degradation
Implement an equivalent feature on Android with the ML Kit GenAI APIs and Gemini Nano, handling model download and unsupported devices
Convert, quantize and integrate a custom model with Core ML on iOS and LiteRT on Android, and choose when ExecuTorch or ONNX Runtime fits better
Profile memory, latency, battery and thermal behavior on physical devices and set release budgets for app size and model footprint
Design a hybrid on-device and cloud fallback with consistent prompts, schemas and privacy disclosures
Syllabus
5 modules · 16 lessons · about 14 hours
1
Module 1: On-Device AI: When and Why
The trade-offs that decide where inference runs, and the hardware that makes it possible.3 lessons
On-Device, Cloud or Hybrid: A Decision Framework for Mobile Features
Neural Engines, NPUs and Unified Memory: What Phones Can Actually Run
Device Coverage: Designing for the Phones That Cannot Run the Model
2
Module 2: iOS: Apple Foundation Models and Core ML
Using the system model and bringing your own.4 lessons
The Foundation Models Framework: Sessions, Availability and Context Limits
Guided Generation with @Generable and Tool Calling in Swift
Core ML and Core ML Tools: Converting and Compressing Your Own Model
MLX Swift for Prototyping on Apple Silicon
3
Module 3: Android: ML Kit GenAI, Gemini Nano and LiteRT
Using the system model through AICore and running custom models with Google's runtime.3 lessons
ML Kit GenAI APIs on AICore: Summarize, Rewrite, Proofread and Prompt
Feature Download, Device Eligibility and Fallback in Kotlin
LiteRT and LiteRT-LM: Running Your Own Models with GPU and NPU Delegates
4
Module 4: Cross-Platform Runtimes and Model Preparation
Shared models across both apps, and the conversion and quantization work behind them.3 lessons
ExecuTorch and ONNX Runtime Mobile: One Model, Two Apps
Quantization for Phones: Weight-Only, Integer and Palettization Trade-offs
Packaging Models: On-Demand Download, Versioning and App-Size Budgets
5
Module 5: Performance, Hybrid Fallback and Privacy
Making it fast, making it work everywhere, and making it trustworthy.3 lessons
Profiling on Real Devices: Latency, Memory, Battery and Thermals
Hybrid Flows: Falling Back to Cloud Models with Firebase AI Logic or Your Own Backend
Privacy, Consent and App Store Review for AI Features
Exam and certificate
The exam unlocks once you have completed every lesson. It draws 20 questions from a larger bank, spread across the modules, with the answer options shuffled for each attempt. You have 40 minutes and need 80% to pass. After a failed attempt you can try again after 24 hours, up to 3 attempts in total. Feedback is given per module, never as the answer key.
Passing earns a certificate of completion and assessment from the Wirewalk Institute of AI with a unique ID that anyone can check at institute.wirewalk.com/verify. It is not an accredited qualification or a vendor certification. How certification works.
The Wirewalk Institute of AI is independent. This course teaches publicly documented products and is not affiliated with or endorsed by Apple (Foundation Models framework, Core ML, Core ML Tools, MLX Swift), Google (ML Kit GenAI APIs, Gemini Nano, AICore, LiteRT, LiteRT-LM, Firebase AI Logic), Meta (ExecuTorch), or Microsoft (ONNX Runtime Mobile). Product and company names are trademarks of their respective owners.
A free, one-hour, hands-on start built on one realistic job: sorting a property manager's 1,150 monthly maintenance requests. You get an API key the safe way, make real calls with curl and an official Python SDK, run them over forty work orders, read stop reasons, usage and errors, estimate the monthly cost, and decide whether the script is ready. Exercises run on your own provider account for well under a dollar.
A code-first course on integrating LLMs into Python the way production code needs it: uv projects and secrets, the OpenAI, Anthropic and Google GenAI SDKs side by side, async concurrency and streaming, Pydantic structured outputs and tools, retries, local models, embeddings and RAG, FastAPI and pandas integration, testing and evals with pytest, observability, cost tracking and deployment.
Call OpenAI, Anthropic, Gemini and local Ollama models from Perl with real CPAN modules, add streaming, concurrency and validated JSON output to text pipelines, and use AI coding assistants to modernize legacy Perl behind characterization tests.