Building the future with high-performance Rust libraries for scientific computing, artificial intelligence, and quantum computing
Leading the Rust ecosystem in scientific and advanced computing
We develop enterprise-grade, open-source software libraries in Rust, bringing memory safety, performance, and fearless concurrency to scientific computing, AI, and quantum computing domains.
With over 32 million source lines of code across 75+ production-grade projects comprising 850 crates and 350,000+ passing tests, we maintain one of the largest open-source Rust ecosystems focused on advanced computing solutions.
Tetsuya Kitahata, CEO and principal developer, leads our engineering efforts. All projects are publicly available on GitHub under permissive open-source licenses.
Production-ready libraries powering the next generation of computing
Comprehensive scientific computing and AI/ML foundation in Rust. SciPy-compatible API with Rust's performance, safety, and concurrency capabilities.
Rust-native semantic web platform. Alternative to Apache Jena/Fuseki with SPARQL 1.2 and GraphQL support from the same dataset.
High-performance SMT solver in pure Rust. Reimplements Z3 with ~90%+ feature parity at ~25% codebase size. No C/C++ dependencies.
High-performance numerical computing library inspired by NumPy. Foundation for scientific computing in Rust with ergonomic APIs.
Logic-to-tensor compiler transforming logical rules into tensor equations for neural-symbolic AI. Bridges symbolic reasoning with neural networks.
Quantum computing framework with type-safe circuit design. Supports 30+ qubit simulations with SIMD and GPU acceleration.
Pure-Rust neural speech synthesis framework. High-performance Text-to-Speech (TTS) with VITS, HiFi-GAN, and DiffWave vocoder.
Pure Rust geospatial data abstraction library (GDAL replacement). Read and write raster/vector geospatial formats without C dependencies.
Production-grade pure Rust BLAS/LAPACK implementation. Drop-in replacement for Fortran BLAS with competitive performance.
PyTorch-compatible deep learning framework entirely in Rust. Tensor operations with sharding for distributed training.
Sovereign Media Framework β pure Rust reconstruction of both FFmpeg and OpenCV, unified in a single cohesive framework. Patent-free, memory-safe multimedia processing.
scikit-learn reimplemented in Rust. Combines scikit-learn's intuitive API with Rust's performance and memory safety.
Legal statute parsing and simulation framework. Transforms legal documents into machine-verifiable code while preserving judicial discretion boundaries.
99% Rust port of FFTW3. High-performance Fast Fourier Transform library with near-native FFTW3 performance.
Ultra-lightweight S3 gateway for AI/HPC workloads. Minimal overhead S3-compatible object storage proxy.
Four-layer RAG engine with SMT-based logic verification. Combines vector search, draft speculation, logic verification, and knowledge graphs.
Comprehensive optimization library for ML. Gradient-based, evolutionary, and constrained optimization algorithms.
DataFrame library inspired by pandas. Ergonomic data manipulation and analysis with Rust's performance and type safety.
Inter-Planet File RUST System. Decentralized file storage and retrieval with content-addressable architecture.
Rust-native AGSP for continuous signal streams. Adaptive generalized signal processing framework.
Spin dynamics simulation in magnetic and topological materials. Micromagnetic and atomistic spin simulations.
Binary zero-fluff encoder/decoder. Compact binary serialization with minimal overhead.
High-performance Rust implementation of Hugging Face Transformers. 21+ architectures with 1.35x faster BERT inference.
Memory-safe, high-performance Interactive Theorem Prover (ITP) natively integrated into the Rust ecosystem. Inspired by Lean 4, bringing formal verification to Rust.
The Oxidized Archiver β pure Rust implementation of archive and compression formats with core algorithms implemented from scratch.
AST-based analysis tool that automatically refactors large Rust source files (>1000 lines) into well-organized, compilable modules.
Graph-based LLM workflow orchestration platform built in Rust. Compose complex AI applications using DAGs with a type-safe, modular approach.
Production-grade Rust-native tensor stack with generalized contraction, sparse/low-rank mixed execution, and out-of-core processing.
Pure Rust morphological analyzer compatible with MeCab dictionaries (IPADIC format). Japanese text tokenization and analysis.
Pure-Rust parametric human body generator running entirely client-side via WebAssembly/WebGPU or natively. Synthesizes detailed 3D human meshes.
Production-ready, Celery-compatible distributed task queue library for Rust. Binary-level protocol compatibility with Python Celery.
Pure Rust implementation of TensorFlow. Full-featured machine learning framework with Rust's safety and performance.
Pure Rust sub-2-bit LLM inference engine for PrismML Bonsai models. Zero-FFI, zero-C/C++ β runs on CPU (SIMD), Apple Silicon (Metal), and NVIDIA (CUDA) without llama.cpp, BLAS, or any C/Fortran runtime.
Pure Rust replacement for the entire NVIDIA CUDA Toolkit stack β cuBLAS, cuDNN, cuFFT, cuSPARSE, cuSOLVER, cuRAND and beyond in ~320K lines of safe Rust across 37 crates. No CUDA SDK, no nvcc, no C/C++ toolchain required at build time.
Unified pure-Rust physics engine targeting the same domains as Bullet (rigid body), OpenFOAM (CFD), LAMMPS (molecular dynamics), and CalculiX (FEM) β with no C or Fortran dependencies.
Pure Rust re-implementation of Japan's Digital Agency government AI legal system "ζΊε (Genai)" β built with Β₯4.4B budget on Google Cloud β with formal verification via Legalis-RS + OxiZ SMT solver.
Pure Rust implementation of the EML operator eml(x, y) = exp(x) β ln(y), expressing all elementary functions as uniform binary trees from a single operator and the constant 1. Based on arXiv:2603.21852.
Comprehensive solutions across the computing spectrum
Linear algebra, FFT, signal processing, optimization, and numerical integration with SIMD acceleration
Neural networks, transformers, auto-differentiation, and graph processing for production ML
Circuit simulation, quantum algorithms, and hardware integration for quantum applications
SPARQL 1.2, GraphQL, RDF processing with AI-augmented reasoning capabilities
Statute parsing, multi-jurisdiction analysis, and formal verification for legal documents
Satisfiability Modulo Theories solving, interactive theorem proving, proof generation, and formal logic verification
Pure Rust geospatial data abstraction, raster/vector format processing, and coordinate transformations without C dependencies
Neural speech synthesis, adaptive signal processing, and real-time audio/visual stream analysis with cross-platform support
Visit our GitHub organization to see all 75+ projects, 850 crates, and join our community
GitHub: cool-japan COOLJAPAN Blog