A collection of custom scikit-learn compatible transformers, metrics, and models to extend the functionality of machine learning pipelines.

rviz is a 3D visualizer for the Robot Operating System (ROS) framework, allowing developers to see robot data in 3D.

An enterprise-grade AI Gateway framework powered by Azure API Management that provides centralized security, cost control, and observability for managing AI models, tools, and agents.

An advanced Python deobfuscator that combines local LLMs with AST cleaning and pattern matching to help malware analysts and reverse engineers recover readable code.

A Python library for the LandingAI ADE API that parses PDFs and images into structured Markdown and extracts typed data using Pydantic models.

A framework for integrating machine learning into LLVM to replace human-crafted compiler optimization heuristics with trained models.

An open-source PyTorch implementation of AlphaFold 3 for predicting the 3D structures of biomolecular interactions, including proteins, nucleic acids, and ligands.

A high-performance Rust implementation of the Kokoro TTS model providing fast text-to-speech synthesis via a CLI, Rust crate, and OpenAI-compatible API server.

An open source agent engineering platform for collaboratively tracing, evaluating, monitoring, and debugging AI agents.

StoryMem is a multi-shot video generation framework that uses a memory-conditioned diffusion model to create long, coherent narrative videos with consistent characters.

A lightweight, OpenAI-compatible AI SDK designed for minimal bundle size and high portability across browser, edge, and agent tooling.

A tool that compiles design systems into installable Agent Plugins, allowing AI coding agents to write on-brand UI by following a specific set of design tokens and composition rules.

ObjectClear is an object removal model that jointly eliminates target objects and their associated effects, such as shadows, while preserving background consistency.

A framework that uses Repository Planning Graphs (RPG) as a persistent control layer for AI coding agents to prevent requirements drift and manage complex, multi-file codebase generation and editing.

A library implementing Wavelet Convolutions to provide CNNs with large receptive fields, featuring optimized backends for CUDA, Metal, and Triton.

A native macOS AI desktop companion that acts as a unified control plane for multiple AI agents and terminals, integrating system diagnostics and cross-device task management.

IBM Analog Hardware Acceleration Kit (AIHWKit) is a Python/PyTorch library that simulates analog in‑memory AI chips (PCM, RRAM, etc.). It provides analog‑aware layers, optimizers, and a high‑performance CUDA/C++ simulator, plus utilities to convert standard models to analog equivalents. Install via pip or conda, run on CPU/GPU (including AMD ROCm), and use it for research, algorithm prototyping, or education on analog AI hardware.

A Three.js-based Web renderer for 3D Gaussian Splatting that enables high-performance streaming and adaptive LOD for large-scale 3D scenes.

A series of discrete diffusion Large Language Models scaled up to 100B parameters, offering high-speed parallel decoding for text generation.

A collection of Triton kernels for efficient low-bit matrix multiplication that accelerates LLM prefill and decoding performance.

WorldEngine is an open‑source post‑training platform for autonomous driving that automatically discovers rare, safety‑critical scenarios from real logs, reconstructs them with photorealistic 3‑D Gaussian Splatting, and fine‑tunes driving models via RL on the generated roll‑outs. It includes two subsystems—SimEngine for closed‑loop simulation and AlgEngine for training/evaluation—showing large gains on rare‑case benchmarks and a real‑world deployment with zero disengagements.

A GPU-native Structure from Motion (SfM) pipeline that accelerates the process of reconstructing 3D structures from 2D images, including integrated support for 3D Gaussian Splatting.

The Cohere Python SDK is a client library that allows developers to access Cohere's AI models across various platforms including AWS, Azure, GCP, and Oracle OCI.

A versatile coding agent that can run in the terminal, as an IDE plugin via ACP, or non-interactively, allowing AI to plan and execute code changes across various LLM providers.

A PyTorch library for the massively parallel training and inference of machine-learning weather and climate models, supporting scaling across 100+ GPUs.

A collection of audio processing scripts for tasks like vocal separation, loudness matching, and transcription to prepare datasets for AI training.

A high-performance automatic differentiation package for Julia that supports mutation and leverages optimized intermediate representations for efficient derivative calculations.

An interactive machine learning toolkit for segmenting, classifying, tracking, and counting cells and other experimental image data.

MDia is a Python library that formalises LLM intermediate reasoning as reusable “dialect cards” (LSFs) and provides a deterministic eight‑stage pipeline (collect → create → evolve → profile → select → run → validate‑rules → report) for generating, evolving, profiling, routing, and auditing these protocols across heterogeneous models. It supports black‑box LLM APIs, offers several routing plans, includes a 100‑rule bank, and guarantees reproducibility by freezing decisions before test evaluation. The repo includes a toy offline demo that runs without any API keys, extensive documentation, MIT licensing, and a citation to the ICML 2026 poster and arXiv paper.

Evo-1 is a lightweight Vision-Language-Action (VLA) model for robotic manipulation that translates visual and textual instructions into robot actions across various simulations and real hardware.

An AI-powered tool that transforms topics or files into interactive, visual mind maps using LLMs and a custom SVG rendering engine.

An AI-powered platform that automates the creation of novels, including text generation, consistent character illustrations, and video adaptations.

A Python module for building topological neural networks that enables deep learning on topological domains like simplicial complexes for science and engineering.

MineBench is a web‑based benchmark that evaluates LLMs on 3‑D spatial reasoning by having them output voxel coordinates for prompts like “a medieval castle”. It visualises the builds, lets humans compare them head‑to‑head, and ranks models with a Bradley‑Terry system. The platform includes an Arena for pairwise voting, a Sandbox for custom builds, a Gallery of community prompts, and export options (GLB, STL, .vox, .schem). It supports major LLM providers and can be run locally with Node.js, pnpm, and Docker.

A ComfyUI extension that accelerates Diffusion Transformer (DiT) models through intelligent caching, providing 1.4x to 2.0x speedups for image and video generation.

Onyx is an open-source AI platform that serves as an application layer for LLMs, providing advanced capabilities like Agentic RAG, deep research, and custom agents.

A pluggable benchmarking framework for evaluating and comparing AI memory and context systems across different providers using standardized datasets and judge models.

The official Python library for RunPod, providing an SDK to create serverless AI workers and a REST API wrapper to manage GPU cloud pods and endpoints.

A centralized engineering playbook that synchronizes coding standards and instructions across multiple AI coding agents and repositories.

A gallery of interactive AI samples for Windows developers to explore local AI models and export them as C# Visual Studio projects.

A memory-optimized inference library that enables large AI models like Stable Diffusion and LLMs to run on ultra-low-RAM devices by streaming weights and using attention slicing.

UniWM is a unified, memory-augmented world model for visual navigation that integrates egocentric visual foresight and planning within a single multimodal autoregressive backbone.

A framework for video world models that stores 3D scene content as latent tokens to improve generation speed and reduce memory overhead by avoiding repeated RGB rendering.

OmniNWM is a research codebase that builds a panoramic, multi‑modal world model for autonomous‑driving simulation. It jointly generates RGB, semantic, depth, and 3‑D occupancy maps from a vehicle trajectory, uses normalized Plücker ray‑maps for precise control, and provides occupancy‑based dense rewards for closed‑loop policy evaluation. The repo includes installation steps (including a manual patch to `transformers`), data preparation instructions for nuScenes, pretrained checkpoint download links, and scripts for inference, out‑of‑distribution testing, and staged training.

VerseCrafter is a controllable video world model that uses 4D geometric control and a GeoAdapter to provide precise, interpretable control over camera and multi-object motion in realistic videos.

A suite of world foundation models for future state prediction that generates visual simulations from text or video prompts to support Physical AI development.

HY-World 1.5 is a real-time interactive world modeling framework that uses streaming video diffusion to generate geometrically consistent 3D environments based on user input.

An MCP server that connects AI assistants to the Gemini/Antigravity CLI, enabling large-scale codebase analysis and sandboxed code execution.