smile
Statistical Machine Intelligence & Learning Engine
What is smile?
Statistical Machine Intelligence & Learning Engine
Overview of smile
What it is
SMILE (Statistical Machine Intelligence & Learning Engine) is a machine learning and AI platform for the JVM, built on Java 25 with APIs for Kotlin and Scala. It covers model building, embedding algorithms, and serving classical ML, ONNX graphs, and LLMs locally. The project is open source and hosted on GitHub.
Key features
- Hundreds of algorithms for classification, regression, clustering, manifold learning, time series, anomaly detection, NLP, deep learning, and native LLM inference (MiniMax M3, Qwen, Llama).
- Engineered with primitive-backed memory layouts, Panama FFM, and Java Vector API (SIMD) for speed and small footprint (0.2 GB GraalVM native micro-container, boots in 0.4 s, idles at ~80 MiB RAM).
- On-device training and adaptation: trains and updates models (Random Forests, Gradient Boosting, K-Means, DBSCAN) directly on live edge or application data.
- Three pillars: Core Engine (embed in application), SMILE Serve (unified inference runtime with Quarkus, supports classic ML, ONNX, LLM chat, Prometheus metrics), and SMILE Agent Studio (local-first desktop workspace with 9 specialized AI agents, Agent Hub, polyglot IDE).
- Provides exact TreeSHAP for tree ensembles and model-agnostic KernelSHAP for local feature attributions.
- Supports continuous batching, prefix-preserving Radix KV caches, speculative decoding, atomic model hot-swapping, Kubernetes health probes, and Prometheus metrics.
Use cases
Embedding machine intelligence in JVM microservices, branch servers, payment terminals, vehicle gateways, or air-gapped nodes. Suitable for real-time, autonomous decision-making on resource-constrained edge gateways to multi-GPU clusters. Also used for building and serving classical ML models, ONNX models, and LLMs locally without cloud dependencies.
Getting started
See the repository README for installation instructions via Maven, SBT (Scala), Gradle (Kotlin), and native libraries (BLAS/LAPACK). Quick start examples are provided in the README.
License and pricing
The license is not stated in the metadata (NOASSERTION). The project is open source, with commercial licensing available from HALO.AI to remove copyleft and source-disclosure obligations for proprietary deployments.
Based on the project's public README and repository metadata.