Projects

### Overview: This comprehensive data science tutorial is designed for in-depth learning and instructor-led training. The repository contains 1838 GitHub stars, making it a valuable resource for anyone interested in mastering machine learning and data science. ### Key Features: - **Lesson 1: Introduction to Numpy** - Learn the basics of NumPy, a fundamental library for numerical computations in Python. - **Lesson 2: Data Wrangling using Pandas** - Master the techniques for cleaning and manipulating data using Pandas, a powerful data analysis library. - **Lesson 3: Plotting in Python** - Understand how to visualize data effectively using Matplotlib and Seaborn. - **Lesson 4: Linear Models for Regression & Classification** - Explore and implement various linear models for both regression and classification tasks. - **Lesson 5: Preprocessing Data** - Learn preprocessing techniques to prepare data for machine learning models. - **Lesson 6: Decision Trees** - Dive into the world of decision trees, a popular algorithm for both classification and regression. - **Lesson 7: Naive Bayes** - Understand the principles and applications of Naive Bayes classifiers. - **Lesson 8: Composite Estimators** - Discover how to combine multiple models to improve predictive performance. - **Lesson 9: Model Selection and Evaluation** - Learn how to select the best model for a given problem and evaluate its performance. - **Lesson 10: Feature Selection Techniques** - Master

Overview: Mellea is a Python library designed to replace brittle prompts and flaky agents with structured, testable AI workflows. It focuses on writing generative programs, replacing LLM calls with typed functions that enforce schemas and guarantees. Key features: - `@generative` decorator for turning typed Python functions into LLM calls. - Natural-language requirements attached to any call, validated and automatically repaired. - Multiple backends including Ollama, OpenAI, HuggingFace, and more. - Easy integration with legacy codebases using `mify`. - Exposes any generative program as an MCP tool. Tech stack: Python Why it matters: Mellea simplifies the process of writing generative AI programs, ensuring that the output is always reliable and verifiable, without the need for brittle prompts or flaky agents. Repository: generative-computing/mellea GitHub stars: 1838

### Overview: Timefold Solver is an open-source AI solver for Java and Kotlin that optimizes scheduling and routing problems. It is designed to simplify planning optimization, making it accessible to developers without extensive coding knowledge. ### Key Features: - Vehicle Routing Problem (VRP) optimization - Employee rostering - Task assignment - Maintenance scheduling - School timetabling - Cloud optimization - Conference scheduling - Job shop scheduling ### Tech Stack: Java ### Why It Matters: Timefold Solver aims to free the world from wasteful planning by providing a powerful AI constraint solver that can handle complex scheduling and routing problems efficiently. Its open-source nature and community-driven development make it accessible to a wide range of users, including developers, businesses, and researchers.

### Overview: This repository contains interactive machine-learning experiments, featuring Jupyter notebooks and demo pages. Each experiment includes a notebook for training models and a demo page for visualizing model performance. The models used are trained with TensorFlow 2 and Keras, focusing on supervised learning techniques like Multilayer Perceptrons (MLPs) and Neural Networks (NNs). ### Key Features: - 🎨 [Launch ML experiments demo](http://trekhleb.github.io/machine-learning-experiments/) - 🏋️ [Launch ML experiments Jupyter notebooks](https://nbviewer.jupyter.org/github/trekhleb/machine-learning-experiments/tree/master/experiments/) - 📝 [Supervised Machine Learning](#supervised-machine-learning) - [Multilayer Perceptron (MLP) or simple Neural Network (NN)](#multilayer-perceptron-mlp-or-simple-neural-network-) ### Tech Stack: Jupyter Notebook ### Why it Matters: These experiments serve as a sandbox for learning and experimenting with different machine learning approaches, algorithms, and data sets. They are not production-ready code or models, but rather a playground for understanding and practicing machine learning techniques. The models may not perform well due to the nature of the experiments, and there is a place for overfitting/underfitting.

### Overview: The Advanced Literate Machinery project aims to develop systems that can read, think, and create, aiming to surpass human intelligence. This project focuses on OCR, with plans to expand into more advanced capabilities in the future. ### Key features: - **CC-OCR**: A comprehensive OCR benchmark designed for evaluating Large Multimodal Models. - **Platypus**: A specialized model for text reading from various forms, leveraging a unified architecture. - **SceneVTG**: A visual text generator that can produce high-quality text images in the wild. - **WebRPG**: A task focused on generating visual presentations for web pages based on HTML code. - **ProcTag**: A process tagging model for assessing the efficacy of document instructions. ### Tech stack: C++ ### Why it matters: This project is crucial for advancing the field of AI, particularly in the areas of reading and understanding text, and could potentially lead to significant breakthroughs in the future.

Overview: ELI5 is a Claude Code skill designed to explain technical concepts to anyone, from kids to adults, adapting tone, vocabulary, and analogies to match the audience's age, grade level, job role, and relationship. Key features: - Adapt tone, vocabulary, analogies, and framing based on the audience - Automatically adjusts for different age groups, grade levels, job roles, and relationships - Supports a wide range of audiences from ages 5 to 40+ Tech stack: Python Why it matters: ELI5 simplifies complex technical concepts into easily understandable explanations, making it a valuable tool for anyone looking to communicate effectively with diverse audiences.

### Overview: Byzer (former MLSQL) is a low-code, open-sourced and distributed programming language designed for data pipeline, analytics, and AI in a cloud-native way. It simplifies data development by treating everything as a table, making it easier to interact with data without switching between different tools. ### Key features: - **Table-Based Design**: Everything is represented as a table, making it easy to interact with data. - **Built-in Algorithms and Extensions**: Includes built-in algorithms and extensions to simplify data processing tasks. - **Distributed Architecture**: Supports distributed processing across multiple nodes. - **Interactive GUI**: Provides a notebook and workflow interaction for a more intuitive data development experience. ### Tech stack: Scala ### Why it matters: Byzer simplifies the process of data development by treating everything as a table, reducing the need to switch between different tools and making the development process more efficient and intuitive. The language's distributed architecture allows for seamless integration with cloud-native environments, making it ideal for modern data pipeline and analytics workflows.

Overview: 静默活体检测（Silent-Face-Anti-Spoofing）项目旨在提供一种无需用户配合的活体检测技术，适用于安卓平台。通过开源的模型、数据预处理方法和脚本，以及提供APK供测试使用，旨在简化开发和测试过程。 Key features: - 开源的活体模型训练架构 - 数据预处理方法 - 模型训练和测试脚本 - 提供适用于安卓平台的APK供测试使用 - 开源的模型剪枝方法，降低计算量与参数量 - 提供关键指标，包括模型输入、FLOPs、速度、FPR和TPR - 提供测试方法和测试须知 - 提供测试须知，包括测试图片要求、设备型号测试数据和相关资源链接 Tech stack: Python Why it matters: 该技术提供了一种无需用户配合的活体检测方法，适用于各种场景，如银行、机场等，提高了安全性并简化了用户体验。

### Overview: This repository contains a collection of NLTK data packages, including corpora, models, tokenizers, and more, essential for developers working with the Natural Language Toolkit. ### Key features: - **Corpora**: A variety of datasets for text analysis. - **Models**: Pre-trained models for natural language processing tasks. - **Tokenizers**: Tools for splitting text into words or other units. ### Tech stack: Python ### Why it matters: Improves the efficiency and effectiveness of natural language processing projects by providing pre-processed and curated data.

### Overview: Atomic Chat is a local AI app and inference engine that allows you to run open-weight language models locally, making it private and offline on your computer. It is compatible with the OpenAI SDK and supports various platforms including macOS, Windows, and Linux. ### Key Features: - **Local AI App**: Runs open-weight language models locally, providing a private and offline experience. - **Node.js Runtime**: Uses Node.js version 20 or higher for seamless integration. - **MacOS, Windows, Linux**: Available as a universal DMG, x64 setup, and AppImage for desktops, and as an iOS and Android app store download. - **OpenAI API Compatibility**: Compatible with the OpenAI API, making it easy to integrate with existing projects. ### Tech Stack: - **Node.js**: Version 20 or higher. - **TypeScript**: Used for development and packaging. - **Tauri**: Built with Tauri for cross-platform application development. - **OpenAI**: Compatible with the OpenAI API for seamless integration with existing projects. ### Why It Matters: Atomic Chat is designed to be a powerful tool for developers and AI enthusiasts who want to leverage advanced AI capabilities locally without the need for internet connectivity. By running models locally, it ensures privacy and reduces the risk of data breaches, making it ideal for environments where security is a priority. ### Repository: - **GitHub Repository**: [AtomicChat](https://github.com/AtomicBot-ai/Atomic-Chat) - **GitHub Stars**: 1810 - **

Overview: EvalPlus is a comprehensive evaluation framework designed for LLM4Code, featuring HumanEval+, MBPP+, and EvalPerf, providing 80x more tests and 35x more tests compared to existing benchmarks. Key features: - HumanEval+ with 80x more tests - MBPP+ with 35x more tests - EvalPerf for rigorous evaluation Tech stack: Python Why it matters: EvalPlus offers unparalleled rigor in evaluating LLM4Code, ensuring developers receive reliable and comprehensive performance metrics. Repository: evalplus/evalplus GitHub stars: 1833

### Overview: The India AI Ecosystem Building Strategy aims to make India the global leader in inclusive, ethical, and affordable AI solutions, focusing on creating a $5 trillion AI economy by 2035. The strategy addresses India's unique challenges and leverages its strengths to accelerate AI innovation. ### Key Features: - **India AI Ecosystem Strategy** - Foundation framework using the SCALE approach - **Implementation Roadmap** - Detailed 90-day action plan with specific deliverables - **Master Action Plan** - Executive summary with 30-60-90 day implementation guide - **Low-Cost Initiatives** - Innovative approaches for resource-constrained innovation - **Sector-Specific Roadmaps** - Detailed plans for key sectors - **Talent Development Framework** - Creating 1 million AI professionals - **Infrastructure Blueprint** - Distributed architecture design - **Policy & Regulatory Framework** - Innovation-enabling regulations - **Partnerships & Funding Strategy** - Mobilizing ₹25,00,000 crore - **Global Positioning Strategy** - Establishing India as an AI leader - **Visual Summary** - Diagrams and frameworks - **High-Leverage Strategies** - Maximum impact with minimal resources - **Top 10 Actionable Initiatives** - Committee-selected best effort-impact ratio initiatives - **Outreach & Communication** - AI-powered distribution plan to reach 1 million Indians - **Government Brief** - For policy makers and bureaucrats - **Industry Guide** - For business leaders and

### Overview: 本项目提供了一个多因子股票量化框架的开源教程，涵盖了从研究框架到因子、回测与组合优化的学习路线。此外，还收录了学界和业界的经典资料，包括AI + 金融的相关工作，如LLM、Agent和benchmark（评估）等。 ### Key features: - 多因子股票量化框架开源教程 - 学界和业界的经典资料收录 - AI + 金融的相关工作，包括LLM, Agent, benchmark(evaluation)等 - 精选资源地图，涵盖量化数据、回测、因子挖掘、机器学习、强化学习与金融课程的持续整理 ### Tech stack: Python ### Why it matters: 为量化研究者和学习者提供了一个全面的资源库，涵盖了从基础到进阶的多因子量化投资框架，以及最新的AI + 金融研究动态，帮助读者快速入门并深入理解量化金融领域的知识和技术。

### Overview: This repository contains a collection of excellent AI-related books and PDFs, as well as a playground environment for learning and experimenting with various AI models and frameworks. It serves as a valuable resource for developers looking to enhance their knowledge in AI. ### Key Features: - **Books**: A diverse range of AI-related books, including introductory theory, mathematics, data mining, deep learning, and more. - **Mathematics**: A dedicated section for frequently used mathematical symbols. - **Training Grounds**: A playground environment for various AI models, including reinforcement learning, feature selection, machine learning, deep learning, language models, computer vision, meta learning, transfer learning, auto ML, and dimensionality reduction. - **Libraries with Online Books**: A collection of libraries and online resources for machine learning, deep learning, reinforcement learning, and more. ### Tech Stack: Jupyter Notebook ### Why It Matters: This repository serves as a comprehensive resource for developers seeking to deepen their understanding of AI through learning and experimentation. It provides access to a wide range of AI-related books and playground environments, making it an invaluable tool for anyone looking to enhance their skills in AI development.

### Overview: Beam Cloud Beta9 is an ultrafast, serverless AI runtime designed for deploying and scaling AI applications with zero infrastructure overhead. It provides fast cold starts, parallelization, developer-friendly features, and supports GPU acceleration. ### Key Features: - **Fast Cold Starts**: Containers launch in under a second using a custom container runtime, scheduler, and embedded caching. - **Parallelization and Concurrency**: Workloads are distributed to 100s of containers for efficient parallel processing. - **First-Class Developer Experience**: Hot-reloading, webhooks, and scheduled jobs. - **Scale-to-Zero**: Workloads are serverless by default, reducing operational overhead. - **Volume Storage**: Supports mounting distributed storage volumes. - **GPU Support**: Runs on cloud GPUs (e.g., NVIDIA A10G, H100) or brings your own GPUs. ### Tech Stack: Go ### Why it Matters: Beam Cloud Beta9 simplifies the deployment and scaling of AI workloads, making it ideal for developers looking to quickly and efficiently run AI applications without the need for complex infrastructure setup. Its fast cold starts, parallel processing capabilities, and serverless architecture make it particularly useful for real-time and batch processing tasks.

### Overview: Zotero-AI-Butler 是一个专为 Zotero 文献管理软件设计的插件，它利用大模型自动精读论文，生成简洁易懂的笔记，帮助用户高效掌握研究内容。 ### Key features: - 自动扫描和总结论文 - 提供Markdown格式的笔记 - 支持多种语言和格式的论文 - 生成学术海报式图片 - 自动生成文献综述报告 - 按思维导图形式呈现论文结构 - 侧边栏支持LaTeX公式和追问功能 - 支持多文献综述和分类管理 ### Tech stack: TypeScript ### Why it matters: Zotero-AI-Butler 提高了文献阅读效率，帮助用户快速掌握研究内容，节省大量时间。通过自动化处理和智能总结，用户可以轻松回顾和复习研究材料，从而提高学习和研究效率。

### Overview: Grounded Docs MCP Server is an open-source alternative to Context7, Nia, and Ref.Tools, providing a comprehensive documentation index for your AI coding assistant. It fetches documentation from various sources including websites, GitHub repositories, npm, PyPI, and local folders, and serves search over the Model Context Protocol (MCP). ### Key Features: - **Documentation Indexing**: Fetches documentation from multiple sources and stores each library version separately. - **Search Engine**: Combines full-text search with optional vector search through OpenAI, Ollama, Gemini, Vertex AI, AWS Bedrock, Azure OpenAI, or any OpenAI-compatible server. - **Local Storage**: Runs on your machine and keeps its index in a local SQLite database. - **Embedding Support**: Supports vector search through OpenAI, Ollama, Gemini, Vertex AI, AWS Bedrock, Azure OpenAI, or any OpenAI-compatible server. - **Usage Telemetry**: Usage telemetry is enabled by default; you can turn it off in `docs/infrastructure/telemetry.md`. - **Compatibility**: Works with any MCP client, over stdio or Streamable HTTP. ### Tech Stack: - TypeScript ### Why It Matters: Grounded Docs MCP Server offers an open-source alternative to popular documentation tools, providing a robust and customizable solution for indexing and searching documentation for AI coding assistants.

Overview: VectorChord is a PostgreSQL extension designed for efficient storage and fast vector search, offering significant cost savings and scalability. Key features: - Affordable vector search: Host 100M × 768-dimensional vectors → AWS i4i.xlarge ($247/month), 1B × 96-dimensional vectors → i7ie.6xlarge ($2246/month). - Accelerated index build: Index 100 million vectors in just 20 minutes using hierarchical K-means and optimized disk operations. - Smoothly scale up: Control memory growth through dimensionality reduction and sampling. - Seamless integration: Compatible with pgvector data types and syntax, providing optimal defaults without complex parameter tuning. Tech stack: Rust. Why it matters: Enables efficient storage and fast vector search, reducing infrastructure costs and providing significant performance improvements over existing solutions. Repository: supervc-stack/VectorChord Primary language: Rust GitHub stars: 1824 GitHub description: Scalable, fast, and disk-friendly vector search in Postgres, the successor of pgvecto.rs.

### Overview: Stakpak is an open-source agent designed to keep your applications running 24/7, ensuring your code is always up-to-date and secure. ### Key Features: - **Secret Substitution**: The LLM works with your credentials without ever seeing them. - **Warden Guardrails**: Network-level policies block destructive operations before they run. - **DevOps Playbooks Baked-in**: Curated library of DevOps knowledge in Stakpak Rulebooks. ### Tech Stack: Rust ### Why it Matters: Stakpak is built differently, offering secure, reliable, and efficient DevOps automation without the risk of AI agents making mistakes. It ensures your production is always up and running. Repository: stakpak/agent Primary language: Rust GitHub stars: 1827 GitHub description: Ship your code, on autopilot. An open source agent that lives on your machines 24/7 and keeps your apps running. 🦀

Overview: MCP Server para 70 APIs públicas brasileiras, fornecendo acesso a dados governamentais em 533 tools. Key features: - 533 tools covering 15 areas including economy, legislature, transparency, judiciary, election, environment, health, education, public security, procurement, fiscal subnational, aviation, energy, and more. - Datasets with local cache, including SIAPA, TSE, ANP, INEP, ISP-RJ, ANAC, and more. - Cross-referencing with `planejar_consulta`, creating execution plans for multiple APIs. - Batch execution with `executar_lote`, discharging queries in parallel. - Smart discovery with `smart_search`, filtering relevant tools based on context. - Auto-registry, adding features by creating directories. - Async everywhere, using httpx async, Pydantic v2, and rate limiting with backoff. Tech stack: Python Why it matters: MCP Server provides a comprehensive and efficient way to access and utilize government data from Brazil, covering various areas such as economy, legislature, transparency, judiciary, election, environment, health, education, public security, procurement, fiscal subnational, aviation, energy, and more. The local cache and batch execution features make it highly useful for data-driven applications and research.