A modular PyTorch framework for Deep Reinforcement Learning that allows users to configure and run RL experiments via JSON specs across various Gymnasium environments.

An AI-powered software engineering platform that uses multi-agent systems and knowledge graphs to automate bug fixing, feature implementation, and codebase navigation.

A Python library that abstracts infrastructure management, allowing AI/ML and scientific research code to run on any cloud or on-prem cluster with a single line of code.

An MCP server that enables AI IDEs to safely execute SQL, manage schemas, and handle authentication for Supabase projects.

A Python package for pool-based active learning that provides various query strategies and a meta-algorithm to automatically select the best strategy.

A lightweight desktop application that lets Claude Desktop users manage and switch between third-party Anthropic-compatible API providers via a local gateway.

An evidence-first, local memory runtime for AI agents that prevents noise ingestion and provides verifiable, bitemporal records with retrieval receipts.

An engineering harness for Claude Code that adds multi-model routing, TDD enforcement, and persistent architectural memory to create a self-reviewing AI development system.

A library for processing sequential data that provides a unified framework for task-based datasets, preprocessing, and evaluation for sequence models.

A transparent HTTP proxy that captures Agent-LLM traffic to provide automated grading, behavioral diagnosis, and source-code probing for AI agent evaluation.

A JavaScript/TypeScript client library that enables Node.js, Deno, and Bun applications to use the pgvector PostgreSQL extension for storing and querying vector embeddings.

A Neovim plugin that integrates the Aider AI pair programmer, enabling users to send buffers, diagnostics, and files to the AI directly from their editor.

A research repository implementing Canon layers for LLMs and linear models to improve architecture design and knowledge storage, including synthetic data generators and pretraining recipes.

A Python package for GFlowNets that provides a decoupled framework for prototyping algorithms, including various environments, loss functions, and sampling strategies.

A local-first dashboard that visualizes Claude Code session data, providing analytics on costs, token usage, and tool activity from local JSONL files.

An agentic telemetry system that provides durable storage and OpenTelemetry instrumentation for content-addressable LLM interactions.

A harness engineering layer for Codex and Claude Code that turns AI coding into a gated workflow with specialized roles for planning, development, and auditing.

A deep learning library for X-Risk optimization that allows users to directly optimize performance measures like AUROC, AUPRC, and NDCG for imbalanced data and ranking tasks.

A visual PyTorch pipeline editor that enables users to build, train, and run image classification models through a drag-and-drop interface without writing code.

An R implementation of the Uniform Manifold Approximation and Projection (UMAP) method for dimensionality reduction of high-dimensional data.

A self-hosted web API that provides free, unlimited access to various LLM providers through a single HTTP interface, eliminating the need for paid API keys.

A power CLI and workflow manager that allows users to interact with LLMs from the terminal and integrate them into automated workflows via Ansible Playbooks.

A GPU and parallel-computing library for C++ that provides a high-level, STL-like interface over the OpenCL API.

A library for preprocessing data with TensorFlow that supports full-pass operations and exports transformations as a graph to prevent training-serving skew.

A library for recording and retrieving metadata associated with machine learning developer and data scientist workflows.

AAABench is a benchmark that gives an AI coding agent a real Unreal Engine 5 editor, a demanding brief to build an open-world game, and hours of unsupervised time — then measures whether the agent can design, build, and self-correct a believable world without any human help.

A JavaScript and TypeScript SDK for building Astrid capsules, enabling developers to create portable Wasm components using familiar JS idioms and decorators.

XGo is a natural-language-style programming language that simplifies Go's syntax and integrates C/C++, Python, and JavaScript/TypeScript libraries into a unified ecosystem.

Algernon is a single‑binary Go web server that supports HTTP/2, HTTP/3, Lua/Teal scripting, many template languages (Markdown, Pongo2, Amber, JSX/TSX, etc.), built‑in CSS preprocessors, multiple database back‑ends, rate limiting, auto‑refresh, a plugin system, and optional Ollama LLM calls—all without external dependencies.

A multi-language software supply chain platform for AI R&D that integrates traditional package repositories with AI model hubs and AI agent-driven management.

A pure Rust implementation of TFHE that enables boolean and integer arithmetics to be performed directly on encrypted data without decryption.

TabPy is a Python server that extends Tableau's capabilities by allowing users to execute Python scripts and functions via table calculations.

A scientific data management platform on AWS that turns cloud data into versioned, context-rich packages to ensure data is findable and reusable for AI and research.

A high-performance dataframe and series library for Elixir that enables fast data exploration and manipulation using a Polars-backed backend.

A Python package for streaming, visualizing, and recording EEG and sensor data from Muse headbands using the Lab Streaming Layer (LSL) protocol.

A Vue.js UI component library designed specifically for building AI-powered application interfaces.

A Python package for implementing optimized stencil computations from symbolic specifications, enabling high-performance execution on CPUs, GPUs, and clusters.

A tidy, unified interface for machine learning models in R that allows users to switch between different computational engines and implementations without changing their syntax.

A comprehensive integration library and CLI tool for 1C:Enterprise and OneScript that provides ready-to-use methods for interacting with over 30 external APIs and services.

An R package that provides a comprehensive set of functions to analyze and describe posterior distributions from Bayesian models, focusing on centrality, uncertainty, and effect significance.

A Caddy web server plugin that blocks or manipulates requests from AI services and cloud providers to prevent unauthorized data scraping for AI training.

A developer agent harness for ClickHouse and real-time analytical infrastructure that provides AI coding agents with the interfaces and context needed to build OLAP applications.

A parametric design tool that automatically generates custom 3D-printable or CNC-cut cases for Single Board Computers based on hardware metadata.

A SQL engine for ad-hoc querying and investigation of files, logs, and system processes without the need for data ingestion or intermediate storage.

A Python implementation of the R tsfeatures package that calculates a wide variety of of summary features from time series data for analysis and meta-learning.

An Emscripten system library that provides a 1:1 C/C++ mapping to the browser's WebGPU API, enabling high-performance, low-overhead graphics and compute in WebAssembly.

A pure Python package for computing Bayesian posteriors and evidences using Dynamic Nested Sampling.

A collection of AI agent skills and prompt templates that enable AI agents to manage Alibaba Cloud infrastructure and utilize its Model Studio AI services.