Projects

Overview: NVIDIA NeMo Speech is a scalable generative AI framework designed for researchers and developers working on Large Language Models, Multimodal, and Speech AI (Automatic Speech Recognition and Text-to-Speech). Key features: - Supports 40 languages for ASR and TTS - Offers controllable latency and 240-2400 concurrent streams - Built on cache-aware Fastconformer architecture - Supports 3 new languages (Ar, Ko, Pt) for MagpieTTS v2607 - Features high-quality offline and streaming inference in English with punctuation and capitalization support Tech stack: Python Why it matters: NVIDIA NeMo Speech provides researchers and developers with a powerful toolset for building and deploying speech AI models, making it easier to create high-quality ASR and TTS applications.

### Overview: AirSim is an open-source simulator for autonomous vehicles built on Unreal Engine, designed for software-in-the-loop simulation with popular flight controllers and hardware-in-loop with PX4. It supports cross-platform compatibility and is used for AI research and deep learning experiments. ### Key features: - **Cross-Platform Compatibility**: Runs on Windows, Linux, and macOS. - **Unreal Engine Integration**: Built as an Unreal plugin, compatible with Unreal Engine environments. - **PX4 Support**: Supports PX4 flight controllers. - **Hardware-In-Loop**: Exposes APIs for physically and visually realistic simulations with PX4. - **Documentation and API**: Detailed documentation and APIs for programmatically interacting with the simulation. ### Tech stack: Unreal Engine ### Why it matters: AirSim is a valuable platform for researchers and developers looking to experiment with autonomous vehicle technology, providing a comprehensive environment for deep learning, computer vision, and reinforcement learning algorithms.

Overview: Rebuild the object in a reference image as a code-only, procedural, quality-gated, animation-ready Three.js model. Key features: - Token-efficient image-to-3D reconstruction - Procedural approach for precise and efficient modeling - Quality-gated process ensuring only the best results are used Tech stack: Python 3.10+ standard library Why it matters: This tool allows developers to quickly and accurately convert images into Three.js models, making it ideal for applications requiring high-quality, procedural 3D models without the need for photogrammetry or mesh extraction.

### Overview: DocsGPT is a private AI platform designed for agents, assistants, and enterprise search. It allows users to turn their documents, sites, and connected apps into AI agents that provide answers with citations and knowledge. The platform supports multiple models and is fully self-hosted, making it accessible and customizable. ### Key features: - 🤖 **Agents and deep research:** agents with their own prompt, knowledge, tools, and model, plus a research mode for multi-step answers. - 🔀 **Visual workflows:** chain agents, conditions, state, and sandboxed code in a drag-and-drop builder. - 📚 **Knowledge from anywhere:** PDFs, Office files, web pages, audio, and more, kept in sync from Google Drive, SharePoint, Confluence, GitHub, S3, and others, with GraphRAG. - 🔎 **Answers with citations:** provides answers backed by sources, enhancing the credibility of the information. ### Tech stack: Python ### Why it matters: DocsGPT enables users to integrate their documents and applications into AI-powered agents, providing a seamless and customizable solution for agents, assistants, and enterprise search. The platform supports multiple models, making it versatile and adaptable to various needs. The self-hosted nature of the platform ensures privacy and control, allowing users to tailor the solution to their specific requirements.

### Overview: This repository contains a topic-wise curated list of Machine Learning and Deep Learning tutorials, articles, and other resources. It serves as an excellent resource for anyone looking to learn and improve their skills in machine learning and deep learning. ### Key Features: - Curated list of Machine Learning and Deep Learning tutorials, articles, and other resources. - Contains a topic-wise curated list of Machine Learning and Deep Learning tutorials, articles, and other resources. - Contains a topic-wise curated list of Machine Learning and Deep Learning tutorials, articles, and other resources. - Contains a topic-wise curated list of Machine Learning and Deep Learning tutorials, articles, and other resources. - Contains a topic-wise curated list of Machine Learning and Deep Learning tutorials, articles, and other resources. - Contains a topic-wise curated list of Machine Learning and Deep Learning tutorials, articles, and other resources. - Contains a topic-wise curated list of Machine Learning and Deep Learning tutorials, articles, and other resources. - Contains a topic-wise curated list of Machine Learning and Deep Learning tutorials, articles, and other resources. - Contains a topic-wise curated list of Machine Learning and Deep Learning tutorials, articles, and other resources. - Contains a topic-wise curated list of Machine Learning and Deep Learning tutorials, articles, and other resources. - Contains a topic-wise curated list of Machine Learning and Deep Learning tutorials, articles, and other resources. - Contains a topic-wise curated list of Machine Learning and Deep Learning tutorials, articles, and other resources. - Contains a topic-wise curated list of Machine Learning and Deep Learning tutorials, articles, and other resources

### Overview: LangBot is a production-grade platform for building AI-powered instant messaging bots, integrating with various chat platforms like Discord, Telegram, Slack, LINE, QQ, WeChat, and more. ### Key Features: - **AI Conversations & Agents**: Multi-turn dialogues, tool calling, and multi-modal support. - **Universal IM Platform Support**: One codebase for Discord, Telegram, Slack, LINE, QQ, WeChat, WeCom, Lark, DingTalk, KOOK. - **Production-Ready**: Access control, rate limiting, sensitive word filtering, comprehensive monitoring, and exception handling. - **Plugin Ecosystem**: Hundreds of plugins, event-driven architecture, component extensions, and MCP protocol support. - **Web Management Panel**: Configure, manage, and monitor bots through an intuitive browser interface. - **Multi-Pipeline Architecture**: Different bots for different scenarios with comprehensive monitoring and exception handling. ### Tech Stack: Python ### Why it Matters: LangBot is trusted by enterprises due to its production-grade features and ease of use, making it the go-to solution for building intelligent AI bots across various platforms.

Overview: This book, titled "Mathematical Foundations of Reinforcement Learning," aims to provide a mathematical introduction to the fundamental concepts, basic problems, and classic algorithms in reinforcement learning. It is designed for senior undergraduate students, graduate students, researchers, and practitioners interested in reinforcement learning. Key features: - Mathematical approach to reinforcement learning - Controlled depth of mathematics to ensure friendliness - Illustrative examples based on grid world tasks - Algorithmic core ideas explained without distractions - Correlated and coherently organized chapters Tech stack: MATLAB Why it matters: This book offers a comprehensive yet accessible introduction to reinforcement learning, making it suitable for students and professionals alike. The mathematical approach ensures a deep understanding of the concepts without overwhelming complexity, while the practical examples and clear explanations make the material engaging and easy to follow.

### Overview: Megatron-LM and Megatron Core are open-source libraries for training transformer models at scale. Megatron-LM is a reference example that includes pre-configured training scripts, ideal for research teams, while Megatron Core is a composable library with GPU-optimized building blocks for custom training frameworks. Megatron Bridge provides bidirectional Hugging Face ↔ Megatron checkpoint conversion with production-ready recipes. ### Key Features: - **Megatron-LM**: A reference example with pre-configured training scripts. - **Megatron Core**: Composable library with GPU-optimized building blocks for custom training frameworks. - **Megatron Bridge**: Bidirectional Hugging Face ↔ Megatron checkpoint conversion with production-ready recipes. ### Tech Stack: Python ### Why it Matters: Megatron-LM and Megatron Core are essential tools for researchers and developers looking to scale transformer model training at scale, offering pre-configured scripts and GPU-optimized building blocks for custom training frameworks.

Overview: Toonflow is an open-source AI video creation platform designed for short dramas and motion comics. It offers a seamless workflow with features like infinite canvas, AI agents, and image/video generation, making it a powerful tool for creative storytelling. Key features: - AI-driven short drama and motion comic creation - Infinite canvas for endless creativity - Integration with various models and tools - Self-hosted and extensible architecture - Cross-platform client and plugin support Tech stack: TypeScript, React, Node.js, and others. Why it matters: Toonflow simplifies the process of creating short videos and animations, allowing users to focus on their ideas rather than technical details. Its AI-driven capabilities and customizable tools make it a versatile platform for both beginners and experienced creators. Repository: HBAI-Ltd/Toonflow-app Primary language: TypeScript GitHub stars: 18,045

### Overview: This repository contains a collection of state-of-the-art papers and datasets related to multimodal large language models (MLLMs). It serves as a comprehensive resource for researchers and developers interested in the latest advancements in the field. ### Key Features: - **MME-Survey**: A comprehensive survey on the evaluation of multimodal language models. - **MME-Survey**: A survey on the evaluation of multimodal language models. - **MME**: A comprehensive evaluation benchmark for multimodal large language models. - **Video-MME**: A comprehensive evaluation benchmark for multimodal large language models in video analysis. - **VITA**: Towards GPT-4o level real-time vision and speech interaction. - **VITA-Audio**: Fast interleaved cross-modal token generation for efficient large speech-language model. - **VITA-1.5**: Towards GPT-4o level real-time vision and speech interaction. ### Tech Stack: The repository uses Python, Jupyter Notebooks, and various datasets and tools for evaluation and benchmarking. ### Why it Matters: This repository is crucial for researchers and developers looking to stay updated with the latest advancements in multimodal large language models. It provides a comprehensive overview of the state-of-the-art in the field, including surveys, benchmarks, and datasets, making it an invaluable resource for anyone interested in this area.

Overview: Rowboat is an AI-powered personal assistant that indexes your work into a living knowledge graph, making it easier to collaborate and get things done. It includes built-in email, notes, browser, code mode, meeting note taker, and project-specific workspaces. Key features: - Built-in email client - Notes - Browser - Code mode - Meeting note taker - Project-specific workspaces Tech stack: TypeScript Why it matters: Rowboat streamlines your workflow by organizing your work into a living knowledge graph, making it easier to collaborate and get things done efficiently. GitHub stars: 180,310 Repository: rowboatlabs/rowboat

Overview: torchvision is a collection of datasets, model architectures, and common image transformations used in computer vision projects. Key features: - Datasets - Model architectures - Image transformations Tech stack: Python Why it matters: torchvision simplifies the process of working with computer vision datasets, models, and transformations, making it easier for developers to experiment with and apply state-of-the-art techniques in image recognition and analysis.

### Overview: CAMEL is an open-source multi-agent framework designed to help researchers explore the scaling laws of agents. This project aims to provide a comprehensive platform for studying and understanding the behavior, capabilities, and risks of agents across various tasks and environments. ### Key Features: - **Data Generation**: Generates realistic data for training and testing multi-agent systems. - **Task Automation**: Automates tasks to streamline research workflows. - **World Simulation**: Simulates complex environments to test agent behaviors. ### Tech Stack: Python ### Why It Matters: CAMEL offers a robust and scalable environment for researchers to study the scaling laws of agents, providing valuable insights into their behaviors and capabilities. The framework's ability to handle large-scale simulations and interactions makes it an essential tool for advancing the field of artificial intelligence and robotics.

Overview: This repository contains resources and materials for AI engineers, including the book "AI Engineering" by Chip Huyen, 2025. It includes various resources such as study notes, chapter summaries, and case studies, as well as tools like a heatmap generator for chatbots. Key features: - Study notes - Chapter summaries - Case studies - Prompt examples - ChatGPT and Claude conversation heatmap generator - Fun tools Tech stack: Jupyter Notebook Why it matters: This book provides a comprehensive guide for adapting foundation models to solve real-world problems, covering both LLMs and LMMs. It includes practical questions and answers to help engineers evaluate and improve their AI applications. Repository: chiphuyen/aie-book GitHub stars: 17,902 GitHub description: [WIP] Resources for AI engineers. Also contains supporting materials for the book AI Engineering (Chip Huyen, 2025) The book is available on Amazon, O'Reilly, and most technical book stores. It's also available in multiple languages, including Japanese, Korean, Chinese, Vietnamese, Spanish, Italian, Portuguese, German, Thai, Russian, Polish, and Ukrainian.

### Overview: WrenAI is an open-source GenBI engine that turns business questions into correct SQL, creating shareable dashboards and keeping business definitions in Git. It supports 22+ data sources and is governed by semantic layers and AI context. ### Key Features: - **Generative BI Engine**: Converts business questions into SQL. - **Governed Semantic Layer**: Ensures correct data interpretation. - **AI Context Layer**: Integrates tribal knowledge and examples. - **22+ Data Sources**: Supports BigQuery, Snowflake, PostgreSQL, ClickHouse, Amazon Redshift, Databricks, and more. ### Tech Stack: Python ### Why it Matters: WrenAI streamlines the process of turning natural-language questions into trusted dashboards and charts, ensuring data accuracy and compliance. It's particularly useful for businesses looking to automate BI processes and maintain data governance. Repository: Canner/WrenAI GitHub Stars: 17,888

### Overview: 此项目是机器学习、深度学习、NLP面试中常考到的知识点和代码实现，也是作为一个算法工程师必会的理论基础知识。项目以各个模块为切入点，帮助大家构建清晰的知识体系。 ### Key features: - 线性回归 - 逻辑回归 - 决策树 - 随机森林 - 梯度提升决策树 - XGBoost - LightGBM - 支持向量机 - 贝叶斯网络 - 马尔科夫 - 主题模型 - 最大期望算法 - 聚类 - ML特征工程和优化方法 - K近邻算法 ### Tech stack: Jupyter Notebook ### Why it matters: 此项目是机器学习、深度学习、NLP面试中常考到的知识点和代码实现，也是作为一个算法工程师必会的理论基础知识。通过此项目，您可以系统地学习和实践这些知识点，提高您的面试和实际工作中应用这些技术的能力。

Overview: This repository offers comprehensive resources and samples for developing and managing generative AI workflows using Google Cloud's Generative AI platform. It includes notebooks, code snippets, and tutorials on various aspects of AI development, such as using Gemini Enterprise Agent Platform, Agent Search, Retrieval Augmented Generation, and more. Key features: - Gemini Enterprise Agent Platform - Agent Search - Retrieval Augmented Generation (RAG) - Vision - Audio - Setup Environment Tech stack: Jupyter Notebook Why it matters: This repository provides a practical guide for developers to understand and implement generative AI solutions on Google Cloud, making it easier to build and manage AI-driven applications efficiently.

### Overview: Janus-Pro is an advanced version of the Janus model series, focusing on improved multimodal understanding and enhanced text-to-image generation capabilities. ### Key Features: - **Optimized Training Strategy**: Enhances model performance through improved training techniques. - **Expanded Training Data**: Increases the amount of data used for training, leading to better model generalization. - **Scaling to Larger Model Size**: Allows for larger model sizes, which improves the model's ability to handle complex tasks and generate high-quality outputs. ### Tech Stack: Python ### Why It Matters: Janus-Pro significantly enhances the capabilities of the Janus model series, making it more effective for multimodal understanding and text-to-image generation tasks. This improvement is particularly valuable in applications where high-quality visual outputs are required, such as in image generation and instruction-following tasks.

Overview: SuperAGI is a developer-first open-source framework designed to enable rapid and reliable development of autonomous agents. It simplifies the process of building, managing, and running useful AI agents. Key features: - Easy integration with various AI models - Robust agent management capabilities - Seamless deployment across different environments - Customizable agent behaviors - Real-time monitoring and analytics Tech stack: Python Why it matters: SuperAGI provides a comprehensive solution for developers looking to create autonomous agents, making it accessible and reliable for various applications.

### Overview: Leon is your open-source personal AI assistant, designed for practical, privacy-aware, and real-world execution. It supports various modes, native skills, tools, context, memory, and local AI providers, ensuring consistent behavior and privacy. ### Key Features: - **Modes**: `smart`, `controlled`, and `agent`. - **Skills**: Native skills for controlled actions and agent skills for `SKILL.md`-backed workflows. - **Tools**: Real tools for work completion, including desktop and browser interfaces. - **Context**: Layered memory for durable preferences, context, and discussion. - **Memory**: Compact self-model and bounded proactive pulse. - **AI Providers**: Supports local and remote AI providers for balance. ### Tech Stack: TypeScript ### Why It Matters: Leon is designed to stay practical, privacy-aware, and grounded in your real environment, making it suitable for both local and remote AI providers, and ensuring consistent behavior and privacy. Repository: leon-ai/leon Primary language: TypeScript GitHub stars: 17584