A collection of ML software builds and tools optimized for the deprecated AMD GFX906 architecture to enable modern AI workloads on legacy hardware.

A Python frontend and library for ComfyUI that allows users to define and run image generation workflows as Python scripts instead of visual node graphs.

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

An MCP server that enables AI agents to search and manage local Zotero libraries using a high-performance C++ search engine for private, local-first RAG.

An agent-first, renderer-independent visualization runtime that allows AI agents to inspect data, plan, validate, and render interactive charts and diagrams.

A Python toolbox for analyzing and simulating neural time series using digital signal processing techniques to study oscillatory and aperiodic activity.

A macOS SwiftUI app for managing multiple Codex/ChatGPT accounts with smart quota-based switching and local/remote API proxy workflows.

An open-source toolkit that integrates genomic analysis into Apache Spark to enable biobank-scale bioinformatics and large-scale genomic data processing.

A desktop workbench for Codex and ChatGPT users that combines a graphical coding assistant with multi-account and third-party provider management.

A low-latency GPU memory copy library based on NVIDIA GPUDirect RDMA that allows the CPU to directly map and manipulate GPU memory to reduce transfer overhead.

A Python library for time series analysis that provides a centralized set of over 65 features across statistical, temporal, spectral, and fractal domains for machine learning preparation.

A GPU-accelerated homomorphic encryption library for Python implementing Paillier and BFV schemes to enable private computations like encrypted nearest-neighbour search.

A modular, multi-modal transactional database for AI and semantic search that integrates document, graph, and vector storage into a single ACID-compliant system.

A high-performance Scala API for Bayesian inference via Markov Chain Monte Carlo, enabling users to build generative models and infer posterior distributions.

A package for estimating model performance using tidy data principles, providing a consistent interface for calculating classification and regression metrics in R.

A GPU stress-testing tool that detects thermal throttling and performance degradation across NVIDIA GPUs to ensure peak performance for ML workloads.

A Julia package for computing Taylor polynomial expansions in one or more variables, supporting automatic differentiation and integral calculus.

A GitHub Action that uses LLMs to review pull request diffs, featuring a comment budget to prevent reviewer fatigue by only posting the most critical findings.

A community-driven web directory and registry for discovering, reviewing, and installing Model Context Protocol (MCP) servers.

A collection of code and guides for building with open-source models on Together AI, covering agents, fine-tuning, RAG, and multimodal vision.

A cross-platform ML developer toolkit that enables fast, GPU-accelerated inference for models like Whisper and Phi in the browser and native apps.

An application framework for vertical AI SaaS that provides the operational infrastructure—auth, billing, workspaces, and usage metering—allowing developers to plug in their own AI agents.

A proxy worker that allows the Cursor editor to connect to a local Ollama server by providing a public endpoint via Cloudflare tunnels.

A library that compiles trained traditional ML models into tensor computations to accelerate inference using neural network frameworks like PyTorch.

Voyager is a fast, in‑memory approximate nearest‑neighbor library for vector/embedding data, offering Python, Java, and Scala bindings with a shared HNSW‑based index format. It’s production‑tested at Spotify, supports macOS/Linux/Windows (including Apple Silicon), and is Apache 2.0 licensed.

Search and Learn is a Hugging Face toolkit that lets you improve open‑source LLM performance by allocating more inference‑time compute. It ships ready‑to‑run scripts and YAML “recipes” for three search‑based inference methods (Best‑of‑N, beam search, Diverse Verifier Tree Search) and includes a small Python library, tests, and documentation for reproducing the authors’ scaling‑compute experiments.

A PyTorch-based library for EEG signal analysis that provides unified data IO, preprocessing transforms, and deep learning models to accelerate brain-signal research.

NanoGPT Slowrun is a benchmark that evaluates language‑model training when the data is fixed (100 M tokens) but compute is effectively unlimited. Four tracks (15 min, 1 h, 2 h, unlimited) let participants experiment with heavy regularisation, large models, and exotic optimisers. The repo provides data‑prep scripts, training code for each track, and a world‑record table that logs every improvement (e.g., swiglu, U‑Net, meta‑gradients, ensembles). To try it, clone the repo, install the Python requirements, run `prepare_data.py`, then launch training with `torchrun` on an 8‑GPU H100 node. Submit a PR with a lower validation loss to claim a spot on the leaderboard.

A development platform and control plane for the Model Context Protocol (MCP) that simplifies building, deploying, and managing MCP servers in Kubernetes.

An intelligent observability platform that uses AI agents and a specialized data plane to automate root cause analysis and diagnostic workflows for system monitoring.

An enterprise-grade LLM API integration system that provides a unified, OpenAI-compatible interface to access multiple AI model providers.

An MCP server that provides a shared, persistent knowledge base for AI agents using atomic memories and an automatically generated knowledge graph.

A Kubernetes API for orchestrating distributed, stateful AI inference workloads using multi-role collaboration and built-in service discovery.

An extension for Podman Desktop that allows users to download, run, and experiment with local LLMs and AI applications using containers.

Optimum‑Benchmark is a Hugging Face library for measuring inference and training performance (latency, memory, energy, throughput) of Transformers, Diffusers, PEFT, etc., across many back‑ends (PyTorch, ONNX Runtime, OpenVINO, vLLM, TensorRT‑LLM, IPEX, Llama‑Cpp, Py‑TXI) and devices (CPU, CUDA, ROCm, XPU, HPU). It provides a Python API and a Hydra‑based CLI, supports distributed launches, and outputs reproducible reports that can be pushed to the Hugging  Face Hub.

A framework for programmatically building and managing training data using weak supervision to replace manual labeling in machine learning pipelines.

A library for standardized evaluation and comparison of machine learning models, providing a wide range of metrics across NLP and Computer Vision.

A Rust library for local generation of text, image, and sparse vector embeddings and reranking, utilizing ONNX Runtime for high performance.

A proof-of-concept project that enables Large Language Models to run entirely inside a PDF file by compiling llama.cpp to asm.js and embedding the model via base64.

A Go library for exact cosine similarity vector search and indexing, featuring integrated support for local GGUF embeddings and OpenAI-compatible cloud APIs.

A universal notation and Python library for formulating complex tensor operations across frameworks like PyTorch, JAX, and NumPy without manual reshaping.

A library of scikit-learn compatible embeddings for text and vision that allows users to easily integrate pretrained encoders into ML pipelines.

A reliability instrumentation library for agentic AI that monitors semantic failures and behavioral drift using specialized SRE metrics (SLIs) to prevent production outages.

A high-performance Windows screen capture library for Python that provides low-latency, high-FPS frames for AI agents and computer vision workloads.

A simulator-independent Python language for building neuronal network models that can run on multiple simulators and neuromorphic hardware without modification.

A general-purpose tensor library that provides a high-level abstraction for parallel computing across CPUs, GPUs, and other hardware accelerators.

Fidelius is a privacy computing middleware based on Intel SGX that enables secure data collaboration by ensuring original data remains available for computation but invisible to the parties involved.

A high-performance Mahjong algorithm library that uses pre-computed lookup tables for millisecond-level winning-hand detection and AI-driven discard decisions.