Verified project record

haifengl/smile

haifengl/smile is a comprehensive machine learning framework for the JVM that provides algorithms for classification, regression, clustering, NLP, deep learning, and data visualization. It delivers high-performance computing with advanced data structures and idiomatic APIs for Scala and Kotlin.

Project overview

This project stands out for bringing state-of-the-art machine learning, including LibTorch-backed deep learning and LLaMA-3 inference, directly to the JVM ecosystem alongside a built-in OpenAI-compatible REST server.

Project type: Model Runtime · Data Processing
Use cases: Data Analysis
Deployment: Refer to project documentation
License: License pending

Best for

Developers, data teams, and AI engineers working within the JVM ecosystem using Java, Scala, or Kotlin.

Key capabilities

Provides core classification, regression, clustering, manifold learning, and association rules algorithms (e.g., SVM, Random Forest, K-Means, t-SNE, FP-growth).
Offers LibTorch-backed GPU/CPU tensor operations, neural network layers, LLaMA-3 inference, and EfficientNet image classification.
Provides text normalization, sentence/word tokenization, POS tagging, stemming, phrase extraction, relevance ranking (TF-IDF, BM25), and WordNet integration.
Features data structures, math, linear algebra, statistical utilities, DataFrame API, and readers/writers for CSV, JSON, Parquet, Arrow, JDBC, Avro.
Provides Swing-based interactive plots and declarative Vega-Lite charts for browsers and Jupyter.
A Quarkus-based REST inference service with OpenAI-compatible API and SSE streaming.
An agentic desktop IDE for data science using Python or SMILE, along with CLI tools for training, predicting, and serving models.
Models can be serialized to disk and loaded in a production environment or inside a Spark job.

Limitations and risks

SMILE v5+ requires Java 25; v4.x requires Java 21; all previous versions require Java 8.
Several algorithms (manifold learning, Gaussian Process, MLP, some clustering) require manual installation of BLAS and LAPACK native libraries.
Employs a dual license model that may require acquiring a commercial license for OEMs, ISVs, VARs.

Getting started

Install Java 25, add the smile-core dependency via Maven, Gradle, or SBT, load your data, and train a model programmatically using the API.
Requires allocating JVM heap memory (configurable via -J-Xmx30G). Users can interact with the system through the desktop GUI, CLI tools, or programmatic APIs.

Evidence and sources

README: SMILE (Statistical Machine Intelligence & Learning Engine) is a comprehensive, high-performance machine learning framework for the JVM. SMILE v5+ requires **Java 25**; v4.x requir…
Release: 6.2.4
README: | **LLM** | LLaMA-3 inference, tiktoken BPE tokenizer, OpenAI-compatible REST server, SSE chat streaming | | **Deep Learning** | LibTorch/GPU backend, EfficientNet-V2 image classi…
README: Data structures, math, linear algebra, statistical utilities, I/O
README: Classification, regression, clustering, manifold learning, and more

Reviewed at: 2026-07-16T13:26:22Z