Sample applications built with SKaiNET — the Kotlin Multiplatform AI framework. Each example is a self-contained project; most are Compose Multiplatform apps that run on Android, iOS, Desktop and the web from a single Kotlin codebase.
Important
About the name
“SKaiNET” is a working project name chosen early in the project’s life as part of a personal learning and experimentation effort, before any trademark considerations were known.
The name is not intended to reference, infringe, or imply association with any existing trademarks, companies, or products. It is not a commercial brand and is not claimed or assignable to any company or organization that contributors may be affiliated with.
If a naming conflict arises, the project name may be changed in the future.
A tiny neural network that learns to approximate sin(x) — and you can train it
live in the app. Visualises the target curve, the model's prediction, and the
network architecture. One Kotlin codebase running on Android, iOS, Desktop (JVM)
and WebAssembly.
📂 SinusApproximator/ · runs on Android · iOS · Desktop · Wasm
Handwritten-digit recognition: draw a digit and a convolutional network classifies it on the spot. A clean-architecture Compose Multiplatform app that loads a pretrained GGUF model and can also retrain it in-app.
📂 MNISTDemo/ · runs on Android · iOS · Desktop · Wasm
Word-vector arithmetic and nearest-neighbour search over pretrained GloVe
embeddings — king - man + woman ≈ queen, fully offline. A Compose Multiplatform
app that shows how SKaiNET handles plain vector data, not just neural nets.
📂 GloVeEmbeddings/ · runs on Android · iOS · Desktop · Wasm
A Qwen3 LLM playground in the browser — chat, completion, translation, tool calls, a tokenizer view and a transformer explainer that visualises attention and residuals, all running on-device with no server and no API key.
📂 KllamaDemo/ · runs on Android · iOS · Desktop · Web · Server
The same MNIST digit detector as a pure-Java CLI — proof of SKaiNET's first-class Java interop from a Kotlin Multiplatform engine.
Predicted digit: 7
Confidence: 98.3%
📂 MnistJavaDemo/ · command-line · Java 21+
If this is your first SKaiNET example, start with Sinus Approximator — it is the smallest end-to-end story (define a network, train it, see it predict) and runs on every platform. Then try MNIST Demo for a real convolutional model, GloVe Embeddings for vector arithmetic, and Kllama Demo to run a local LLM.
Each example folder has its own README.md with build and run instructions.