The TypeScript Toolkit for
AI & Numerical Computing

Comprehensive, type-safe library for numerical computing, tabular data workflows, and machine learning. Built to be ergonomic, predictable, and strongly typed.

Read the DocsView on GitHub
Quick install:
npm install deepbox
Node.js 24.13.0+
tensor-ops.ts
import { tensor, add, GradTensor } from "deepbox/ndarray"// Tensor operations with broadcastingconst a = tensor([[1, 2], [3, 4]])const b = tensor([[5, 6], [7, 8]])const c = add(a, b)// Automatic differentiationconst x = new GradTensor([2, 3], { requiresGrad: true })const y = x.mul(x).sum()y.backward()console.log(x.grad) // [4, 6]
neural-net.ts
import { Sequential, Linear, ReLU } from "deepbox/nn"import { Adam } from "deepbox/optim"const model = new Sequential(  new Linear(10, 64),  new ReLU(),  new Linear(64, 1))const optimizer = new Adam(model.parameters(), {   lr: 0.001 })

Everything you need for ML in TypeScript

A complete numerical computing and ML toolkit — now with full TypeScript support and type safety.

N-Dimensional Arrays

90+ tensor operations with full broadcasting and multiple dtypes support.

Automatic Differentiation

GradTensor with full backward pass support for neural network training.

DataFrames & Series

Tabular API with 50+ operations for data manipulation and analysis.

Neural Networks

Linear, Conv, RNN/LSTM/GRU, Attention layers with optimizers and schedulers.

Machine Learning

Classical ML models including Trees, SVM, KNN, Naive Bayes, and Ensembles.

Statistics & Metrics

40+ ML metrics, hypothesis tests, correlations, and statistical analysis.

Ready to get started?

Explore the documentation to learn how to build with Deepbox.

Get Started