Projects

### Overview: yomiyasu（よみやす）はAIが生成した日本語の不自然な比喩、曖昧な主述関係、不要な装飾を直し、読みやすい文章に整えるスキルです。主に技術記事、設計書・仕様書、PR説明文、社内レポートなどの実務的な文章を対象としています。 ### Key features: - 文章の読みやすさを改善するための7つの変換原則を適用 - 文の働きと文体の保持、読点とラベルの決まりを加えた2.0.0バージョン - 原稿の種類と読者知識を考慮した書き直しテスト - 数値や条件の補正、仕様書での許可・能力・保証の言い方の検査 ### Tech stack: Python ### Why it matters: このスキルはAI生成の日本語文章を読みやすい形に整え、技術発信や実務的な文書作成に効果を示すことができます。また、AIの不自然さを直すことで、読者に信頼性と正確性を提供し、業務の効率化につながります。

### Overview: PhotoDemon is a free, portable photo editor that offers comprehensive features, high performance, and maximum usability. It is 100% open-source and available under a permissive BSD license. Contributors have translated the program into more than a dozen languages, making it accessible to a global audience. ### Key Features: - **Lightweight and Portable:** No installer is required, and it leaves no trace on your hard drive. - **Integrated Macro Recording and Batch Processing:** Complex editing actions can be recorded as macros, and a built-in batch processor allows applying macros to entire folders of images. - **Usability is Paramount:** Small touches like real-time effect previews, save/load presets on all tools, unlimited Undo/Redo, customizable hotkeys, mouse wheel and X-button support, and descriptive icons make it fast and easy to use. - **Pro-Grade Features and Tools:** Extensive file format support, advanced multi-layer support, color-managed workflow, on-canvas tools, adjustment tools, filters, and effects. - **Limitations:** Only available as a 32-bit application, and not officially supported on macOS, Linux, or BSD systems. ### Tech Stack: - Visual Basic 6.0 ### Why It Matters: PhotoDemon is a versatile and user-friendly photo editor that offers a wide range of features and tools, making it a valuable tool for both beginners and experienced photographers. Its lightweight nature and open-source nature make it accessible to a global audience, and its focus on usability ensures that it is easy to use and navigate.

Overview: RL4LMs is a modular library designed to fine-tune language models to human preferences, providing customizable building blocks for training transformer-based models on various NLP tasks. Key features: - Implementations of on-policy algorithms - Reward functions - Metrics - Datasets - Actor-Critic Policies - NLPO (Natural Language Policy Optimization) Tech stack: Python Why it matters: RL4LMs offers a comprehensive set of tools for fine-tuning language models to optimize arbitrary reward functions on diverse NLP tasks, making it a valuable resource for researchers and practitioners in the field of natural language processing.

Overview: Spektral is a Python library for graph deep learning, built on top of Keras and TensorFlow 2, aiming to simplify the creation of graph neural networks (GNNs) and provide a flexible framework for various graph-based tasks. Key features: - Supports various graph neural network layers, including GCN, Chebyshev convolutions, GraphSAGE, ARMA convolutions, Edge-Conditioned Convolutions, Graph Isomorphism Networks, Diffusional Convolutions, and many more. - Provides pooling layers such as MinCut pooling, DiffPool, Top-K pooling, Self-Attention Graph (SAG) pooling, Global pooling, and Global gated attention pooling. - Includes utilities for graph representation, manipulation, and transformation, making it easier to integrate graphs into graph neural network models. - Compatible with Python 3.6 and above, tested on Ubuntu, MacOS, and Windows, and available on PyPi and Google Colab. Tech stack: Python, Keras, TensorFlow 2. Why it matters: Spektral simplifies the creation and implementation of graph neural networks, making it easier to apply these models to various graph-based tasks such as node classification, link prediction, and more.

### Overview: Agency Orchestrator 是一个一站式解决方案，旨在通过一句话组建你的“一人公司”AI 专家团队。只需几分钟，即可完成从需求分析到交付的完整方案，涵盖网页、桌面和 Docker 等渠道。AI 专家团队由 276 个专业角色组成，支持 YAML 零代码，提供 10 种免 key 方式，包括 11 个关键免费方案。项目已获得 2394 GitHub 星，适用于多种语言环境，确保了极高的可用性和可靠性。 ### Key Features: - **一句话组建AI专家团队**：只需一句话，AI 专家团队自动组队并运行。 - **276个专业AI角色**：涵盖网页、桌面和Docker等多种渠道，提供10种免key方案。 - **零代码YAML配置**：无需编写代码，使用YAML配置即可。 - **10种免key方式**：支持11种关键免费方案。 - **网页/桌面/Docker渠道**：提供网页、桌面和Docker等多种渠道的解决方案。 - **自动组队与决策**：自动组队并进行重大决策，确保方案的高效性和准确性。 - **验收自动核验**：交付的方案通过验收自动核验，确保质量。 ### Tech Stack: - **TypeScript**：项目的主要开发语言，确保代码质量和性能。 ### Why It Matters: Agency Orchestrator 是一个革命性的工具，简化了组建和管理AI专家团队

### Overview: This book provides a structured and comprehensive approach to mastering essential skills needed to recognize and solve complex real-world problems with Machine Learning and Deep Learning. It is packed with over 500 pages of useful information, including concepts, methodologies, hands-on examples, and code, to help readers master the art and science of Machine Learning. ### Key features: - Hands-on examples and code - Real-world case studies leveraging Python Machine Learning ecosystem - Structured and comprehensive three-tiered approach - Over 500 pages of useful information - Learn-by-doing approach - State-of-the-art data analysis, machine learning, and deep learning frameworks ### Tech stack: Python, Jupyter Notebook, scikit-learn, pandas, statsmodels, spaCy, nltk, gensim, TensorFlow, Keras, Skater ### Why it matters: This book is your perfect companion for learning the art and science of Machine Learning to become a successful practitioner. It provides a structured and comprehensive approach to mastering essential skills needed to recognize and solve complex problems with Machine Learning and Deep Learning.

Overview: This repository is a curated list of resources for autonomous vehicles, including foundational materials, courses, papers, research labs, datasets, open-source software, hardware, toys, companies, media, and laws. Key features: - A curated list of awesome Machine Learning frameworks, libraries, and software. - A curated list of awesome computer vision resources. - Courses and specializations in machine learning and deep learning. - A list of research labs in robotics and autonomous vehicles. - Datasets for computer vision and autonomous driving. - Open-source software for autonomous vehicles. - Hardware and toys for autonomous vehicles. - Companies involved in autonomous vehicles. - Media coverage of autonomous vehicles. Tech stack: Python, JavaScript, HTML/CSS, Markdown, and various programming languages. Why it matters: This repository serves as a comprehensive guide for anyone interested in the field of autonomous vehicles, providing a wealth of resources to help beginners and experts alike understand the latest advancements and technologies in the field. GitHub stars: 2,396 Repository: manfreddiaz/awesome-autonomous-vehicles

### Overview: `verl-agent` is an extension of `veRL`, designed for training large language model (LLM) agents via reinforcement learning (RL). It introduces a step-independent multi-turn rollout mechanism, enabling highly scalable training for tasks requiring up to 50 steps. ### Key Features: - **Step-Independent Multi-Turn Rollout Mechanism**: Allows for fully customizable per-step input structures, history management, and memory modules. - **Rich Suite of RL Algorithms**: Supports algorithms like `GraphGPO` and `HGPO`. - **Diverse Set of Agent Environments**: Enables development of reasoning agents in both visual and text-based tasks. ### Tech Stack: Python ### Why It Matters: `verl-agent` is essential for training large language model agents via reinforcement learning, offering a scalable and customizable approach to complex RL tasks.

### Overview: VCP (Variable & Command Protocol) 是一个面向 AGI OS 开发和探索的工业级基建示范项目。通过统一指令协议、多层级持久化记忆、分布式插件引擎及多 Agent 协作框架，将原本“无状态、无记忆、无工具调用能力”的大语言模型，彻底改造成拥有永久自我意识、物理世界操作权及群体协作智能的完整智能体系统。 ### Key features: 1. **统一指令协议**：确保所有交互指令在系统内部统一处理，减少歧义。 2. **多层级持久化记忆**：系统能够存储和检索多层级的用户和模型数据，提供更丰富的上下文信息。 3. **分布式插件引擎**：支持第三方插件的集成，扩展系统的功能和灵活性。 4. **多 Agent 协作框架**：通过多 Agent 结构实现更复杂的协作和决策过程。 ### Tech stack: - **JavaScript**：作为主要编程语言，实现系统的逻辑和交互。 - **Rust**：用于底层的性能优化和数据处理，确保系统的高效性和稳定性。 ### Why it matters: VCP 的设计目标是将 AI 转化为一个能够持续存在、具备物理世界操作权和群体协作智能的完整智能体系统。通过这些技术手段，VCP 实现了对传统 AI 模型的彻底改造，使其具备了与人类相似的感知、记忆和行为能力，从而在实际应用中展现出显著的优势。

### Overview: Texar is a toolkit designed for machine learning, natural language processing, and text generation tasks, providing a library of easy-to-use ML modules and functionalities for composing models and algorithms. It is compatible with TensorFlow and PyTorch, and is actively contributed by Petuum and Carnegie Mellon University. ### Key Features: - **Two Versions, Mostly Same Interfaces**: Texar-TensorFlow and Texar-PyTorch have mostly the same interfaces, combining the best design of TensorFlow and PyTorch. - **Rich Pre-trained Models**: Supports BERT, GPT2, XLNet, and other pre-trained models for encoding, classification, generation, and composing complex models. - **Fully Customizable**: Supports both novice-friendly and expert-friendly interfaces, allowing easy integration of external modules. - **Versatile**: Supports a wide range of tasks, models, algorithms, data processing, evaluation, and more. - **Modularized**: Based on principled decomposition of Learning-Inference-Model Architecture, providing maximal re-use and clean APIs. - **Distributed Training**: Supports model training with multiple GPUs. - **Documentation and Examples**: Rich documentation and examples available at https://texar.readthedocs.io/. ### Tech Stack: Python ### Why it Matters: Texar is essential for researchers and practitioners looking to fast prototype and experiment with machine learning, natural language processing, and text generation tasks. Its modular and customizable design makes it suitable for both beginners and advanced users.

Overview: Hedgehog Lab is an open-source scientific computing environment in the browser, designed for JavaScript-based scientific computing and data visualization. It supports matrix operations with GPU acceleration, TeX support, and data visualization and symbolic computation. Key features: - Matrix operations with GPU acceleration - TeX support - Data visualization - Symbolic computation Tech stack: TypeScript Why it matters: Hedgehog Lab offers a browser-based solution for scientific computing and data visualization, providing a powerful tool for JavaScript enthusiasts and researchers. Its GPU-accelerated matrix operations and TeX support make it an ideal choice for scientific and technical applications. Repository: Hedgehog-Computing/hedgehog-lab Primary language: TypeScript GitHub stars: 2391 GitHub description: Run, compile and execute JavaScript for Scientific Computing and Data Visualization TOTALLY TOTALLY TOTALLY in your BROWSER! An open source scientific computing environment for JavaScript TOTALLY in your browser, matrix operations with GPU acceleration, TeX support, data visualization and symbolic computation.

Overview: This project provides a provider-neutral Agent Skill designed for various applications beyond coding agents, including research, support, operations, sales, finance, data analysis, procurement, legal workflows, healthcare workflows, education, and workflow automation agents. Key features: - Generates an MVP blueprints for agent harnesses - Validates, authorizes, executes, records, and returns observations - Applies to multiple domains and workflows Tech stack: Skills Why it matters: The skill ensures consistent runtime discipline across different domains and workflows, promoting best practices and standardization. Repository: DenisSergeevitch/agents-best-practices GitHub stars: 2389 GitHub description: Provider-neutral Agent Skill for Codex, Claude Code, and agentic harness design.

### Overview: Webclaw is a Rust-based library and API designed for extracting structured data from websites, specifically for use with large language models (LLMs). It provides a fast, local-first approach to web content extraction, allowing developers to easily integrate clean, structured data into their AI applications. ### Key Features: - **CLI**: Command-line interface for easy integration. - **MCP Server**: Self-hostable server for more advanced use cases. - **REST API**: RESTful API for seamless integration with other systems. - **Structured Data Extraction**: Extracts structured data directly from the web, making it ready for use with LLMs. - **Contextual Data**: Provides contextually relevant data, enhancing the quality of the extracted information. ### Tech Stack: - **Rust**: The primary programming language used for the library and API. - **CLI**: Command-line interface for easy integration. - **MCP Server**: Self-hostable server for more advanced use cases. - **REST API**: Provides a RESTful API for seamless integration with other systems. - **Structured Data Extraction**: Extracts structured data directly from the web, making it ready for use with LLMs. - **Contextual Data**: Provides contextually relevant data, enhancing the quality of the extracted information. ### Why It Matters: Webclaw simplifies the process of extracting structured data from websites, making it easier for developers to integrate clean, structured data into their AI applications. This tool is particularly useful for developers working with large language models, as it ensures that the extracted data is contextually relevant and ready

### Overview: agentUniverse is a powerful multi-agent framework built on large language models, designed to help developers easily create intelligent agents that collaborate effectively in various fields. ### Key features: - **Flexible Multi-Agent Collaboration**: Supports multiple agents with different roles, enabling complex problem-solving. - **Domain Experience Integration**: Seamlessly integrates domain-specific knowledge to enhance agent capabilities. - **Rich Pattern Components**: Offers a variety of pattern components for multi-agent collaboration, including PEER and DOE. ### Tech stack: Python ### Why it matters: agentUniverse simplifies the process of building intelligent agents, making it easier for developers to integrate domain expertise and collaborate effectively in real-world applications.

Overview: OpenNMT is an open-source neural machine translation system built on the Torch framework, designed for efficiency and state-of-the-art translation accuracy. Key features: - Speed and memory optimizations for GPU training - Simple, general-purpose interface - C++ implementation for easy deployment - Extensions for summarization and image captioning Tech stack: Lua Why it matters: OpenNMT provides a robust, efficient, and user-friendly solution for neural machine translation, suitable for both academic and practical applications. Repository: OpenNMT/OpenNMT GitHub stars: 2386 GitHub description: Open Source Neural Machine Translation in Torch (deprecated) [](https://travis-ci.org/OpenNMT/OpenNMT) [](https://codecov.io/gh/OpenNMT/OpenNMT)

Overview: PawWork is a Chrome extension that allows users to interact with the currently logged-in Chrome environment programmatically, enabling them to execute JavaScript, take screenshots, and manage session artifacts. This feature is designed to be BYOK (Bring Your Own Key) and sandboxed, offering a unique way to interact with the browser without relying on the Chrome Web Store. Key features: - Execute guest JavaScript against the browser - Keep session artifacts in the session - Unpacked extension for BYOK and sandboxing Tech stack: JavaScript Why it matters: PawWork provides a powerful and flexible way to interact with the Chrome environment, allowing users to automate tasks and gain insights into the browser's state without the need for a hosted model or server, making it ideal for developers and testers who want to leverage the power of AI while maintaining control over their development environment.

### Overview: AI Engineering Academy is a comprehensive resource for anyone looking to master applied AI concepts. It offers structured learning paths, hands-on practice, and industry-aligned content to ensure practical skills are developed. ### Key Features: - Structured Learning Paths: From fundamentals to advanced concepts. - Hands-on Practice: Real-world projects and implementations. - Industry-Aligned Content: Focus on practical, production-ready skills. - Community-Driven: Learn alongside peers and experts. ### Tech Stack: Jupyter Notebook ### Why It Matters: AI Engineering Academy provides a clear and accessible way to master applied AI concepts, making complex ideas practical and relevant for everyone. ### Repository: adithya-s-k/AI-Engineering.academy ### GitHub Stars: 2384 ### GitHub Description: Mastering Applied AI, One Concept at a Time

Overview: This repository contains the official implementations and experiments for the Large Concept Models (LCM) project, focusing on language modeling in a sentence representation space. Key features: - Auto-regressive sentence prediction using MSE regression - Variants of diffusion-based generation - Models operating in a quantized SONAR space Tech stack: Python Why it matters: The LCM project offers a comprehensive solution for language modeling in a sentence representation space, providing insights into large concept models and their applications in various domains. GitHub stars: 2385

### Overview: This repository provides a minimal implementation of Proximal Policy Optimization (PPO) using PyTorch for OpenAI gym environments. It is designed for beginners to understand the PPO algorithm and can be used for complex environments with some hyperparameter tuning. The repository includes features such as continuous action space with linearly decaying standard deviation, simple Monte Carlo advantages, and single-threaded implementation. It also includes utilities for logging, plotting graphs, and making gifs from pre-trained networks. ### Key Features: - Continuous action space with linearly decaying standard deviation - Simple Monte Carlo advantages - Single-threaded implementation - Training and testing scripts - Log files for episodes, timesteps, and rewards - Utilities for plotting graphs and making gifs - Pre-trained network testing and gif generation ### Tech Stack: PyTorch ### Why It Matters: This implementation is suitable for beginners in Reinforcement Learning to understand the PPO algorithm and can be used for complex environments. The repository includes utilities for logging, plotting, and making gifs, making it easier to analyze and visualize the training process. ### GitHub Stars: 2384 ### Primary Language: Python ### Repository: nikhilbarhate99/PPO-PyTorch ### GitHub Description: Minimal implementation of Proximal Policy Optimization (PPO) in PyTorch. Primarily intended for beginners in Reinforcement Learning. Can be used for complex environments but may require some hyperparameter tuning. Includes utilities for logging, plotting, and making gifs. ### Usage: - To train a

### Overview: AI 엔지니어 기술 면접 스터디는 2,000 이상의 GitHub 라이스타를 보유하고 있으며, 다양한 AI 기술에 대한 질문을 다루고 있습니다. 이 스터디는 통계/수학, 머신러닝, 딥 러닝 등 다양한 주제에 대한 질문을 다루고 있으며, 이를 통해 AI 엔지니어로서의 기술 스킬을 향상시키는 데 도움이 될 수 있습니다. ### Key Features: - **Interview Questions**: - **통계/수학**: 고유값과 고유벡터, 샘플링, 리샘플링, 확률 모형, 누적 분포 함수, 확률 밀도 함수, 조건부 확률, 공분산, 상관계수, 신뢰 구간, p-value, R square, 중심극한정리, 모수적 방법론, 비모수적 방법론, likelihood, probability, bootstrap, 모수적 모델, 비모수적 모델, likelihood, probability, bias, 로그 함수, 베르누이 분포, 이항 분포, 카테고리 분포, 다항 분포, 가우시안 정규 분포, t 분포, 카이제곱 분포, F 분포, 베타 분포, 감마 분포, 출장 비행기 비가 내릴 확