Projects

Overview: This repository serves as a log of the skills and knowledge I am continuously acquiring to enhance my software engineering abilities and expand my knowledge across various technologies. Key features: - System Design: Covers foundational concepts in designing machine learning systems and data-intensive applications. - Maths: Includes courses on probability, statistics, hypothesis testing, and statistical thinking. - Data Structures and Algorithms: Focuses on algorithms and data structures, including linear algebra and eigenvectors. Tech stack: None specified. Why it matters: This log helps me stay updated with the latest technologies and methodologies in software engineering, ensuring I remain proficient and adaptable in the ever-evolving field of software development.

### Overview: InterpretML is an open-source package that provides state-of-the-art machine learning interpretability techniques, enabling you to train interpretable models and explain blackbox systems. This package helps you understand your model's behavior and improve feature engineering, detect fairness issues, ensure human-AI cooperation, and meet regulatory compliance and high-risk application requirements. ### Key Features: - **Explainable Boosting Machine (EBM)**: An interpretable model developed at Microsoft Research, which uses modern techniques like bagging, gradient boosting, and automatic interaction detection to enhance traditional GAMs, making them as accurate as random forests and gradient boosted trees. EBMs provide exact explanations and are editable by domain experts. - **Automatic Interaction Detection (AID)**: Identifies interactions between features that are not obvious to the human eye, helping in understanding complex models. - **Feature Engineering**: Helps in improving the model by understanding the reasons behind individual predictions. - **Model Debugging**: Provides insights into why a model made a mistake, aiding in debugging and improving model performance. - **Detecting Fairness Issues**: Helps in identifying and mitigating biases in the model. - **Human-AI Cooperation**: Facilitates better understanding and trust in model decisions, crucial for applications like healthcare and finance. - **Regulatory Compliance**: Ensures that the model complies with legal requirements, essential for high-risk applications. ### Tech Stack: - Python 3.10+ | Linux, Mac, Windows ### Why It Matters: - **Model Debugging**: Understand why a model made a mistake. -

### Overview: MindSearch is an LLM-based Multi-agent Framework designed to mimic human minds in web search. It is inspired by Perplexity.ai Pro and SearchGPT, offering a powerful and efficient search engine solution. ### Key features: - **Multi-agent Search Engine**: Employs multiple agents to enhance search accuracy and speed. - **Concurrency Optimization**: Improves performance by optimizing the agent module for better concurrency. - **User-friendly Interface**: Simultaneously handles multiple queries for a seamless search experience. ### Tech stack: - **Frontend**: React, Gradio, Streamlit - **Backend**: FastAPI, SQLite (for storing search history) - **Search Engines**: DuckDuckGoSearch, BingSearch, BraveSearch, GoogleSearch, TencentSearch ### Why it matters: MindSearch leverages the power of large language models to provide a more intuitive and efficient search experience. By mimicking human cognitive processes, it aims to enhance the accuracy and speed of web searches, making it a valuable tool for developers and researchers in the field of AI and web development. Repository: [InternLM/MindSearch](https://github.com/InternLM/MindSearch) GitHub stars: 6947

### Overview: This repository is a comprehensive reading list for research topics in multimodal machine learning, curated by Paul Liang. It includes tutorials, papers, and datasets from leading researchers in the field. ### Key features: - Tutorials on Multimodal Machine Learning - Course content from 11-777 Multimodal Machine Learning - Reading list for foundational and advanced topics in multimodal machine learning - Papers, slides, and videos from CVPR 2022 and NAACL 2022 - Comprehensive survey papers on principles, challenges, and open questions in multimodal machine learning ### Tech stack: Python, LaTeX, Markdown ### Why it matters: This repository serves as a valuable resource for researchers, students, and practitioners interested in the latest advancements and challenges in multimodal machine learning. It provides a curated collection of papers, tutorials, and datasets that cover a wide range of topics, from foundational principles to cutting-edge applications.

Overview: LLM Scraper is a TypeScript library designed to extract structured data from any webpage using LLMs. Key features: - Supports multiple LLM models (GPT, Sonnet, Gemini, Llama, Qwen) - Uses Zod or JSON Schema for schema definition - Provides type safety with TypeScript - Based on Playwright framework - Supports streaming objects - Offers 6 formatting modes: HTML, Raw HTML, Markdown, Text, Image, Custom Tech stack: TypeScript Why it matters: LLM Scraper simplifies the process of extracting structured data from webpages, making it easier and more efficient for developers to work with LLMs in their projects.

### Overview: TurboFieldfare is a custom Swift + Metal runtime designed to run the Gemma 4 26B-A4B model on any Apple Silicon Mac, including those with 8 GB of RAM. This model can be efficiently processed without loading the entire 14.3 GB model into memory, instead keeping the shared 1.35 GB core and FP16 KV cache in memory and streaming only the experts needed for each token from SSD. The runtime, streaming installer, CLI, and native Mac app are written in Swift and Metal, making it model-specific rather than a wrapper around MLX or llama.cpp. The curated experiment record summarizes 103 measured results across kernels, caching, I/O, prefill, and decode. ### Key features: - Model: Gemma 4 26B-A4B, 26B total parameters, about 3.88B active per token - Weights: MLX affine 4-bit, group 64; 8-bit router; 4-bit shared and routed experts - Memory: ~2 GB of weights and 4K KV cache - Storage: About 14.3 GB for the text model, plus about 1.1 GB for the optional image pack - Hardware: Apple Silicon Mac; 8 GB of RAM - Platform: macOS 26, Metal 4, Swift 6.2 - M2 measured decode: 5.1-6.3 tok/s on an 8 GB M2 MacBook Air

### Overview: 掌握机器学习实战技巧，轻松构建和部署深度学习模型。 ### Key features: - 使用 PyTorch 实现基本的机器学习算法和深度神经网络。 - 在浏览器中使用 Google Colab 运行所有程序。 - 学习产品级的面向对象机器学习编程。 ### Tech stack: Jupyter Notebook ### Why it matters: 通过本项目，你可以快速上手机器学习实战，掌握从数据中获取有价值见解的方法。无论你是初学者还是有一定基础的开发者，都能从中学到实用的技能。

Overview: 飞桨产业级模型库是面向企业研发和快速集成的开源模型库，包含大量经过产业实践打磨的主流模型和在国际竞赛中夺冠的模型，提供端到端的开发套件，服务领域广泛，包括能源、金融、工业、农业等。 Key features: - 多个场景的开发套件，如语义理解、图像分类、目标检测、图像分割、文字识别、语音合成等。 - 飞桨官方模型超过600个，生态模型超过260个，持续更新中。 - 动态图和静态图模型，支持多种开发场景。 - 飞桨官方实现并提供持续技术支持及答疑。 - 与飞桨核心框架版本对齐，经过充分测试。 Tech stack: Python。 Why it matters: 飞桨产业级模型库是面向企业研发和快速集成的开源模型库，提供端到端的开发套件，满足企业低成本开发和快速集成的需求。 Repository: PaddlePaddle/models Primary language: Python GitHub stars: 6931 GitHub description: 官方模型库包含大量经过产业实践打磨的主流模型和在国际竞赛中夺冠的模型，提供面向语义理解、图像分类、目标检测、图像分割、文字识别、语音合成等场景的多个端到端开发套件，服务领域广泛，包括能源、金融、工业、农业等。

Overview: SQLBot 是一款基于大语言模型和 RAG 技术的智能问数系统，旨在提供对话式数据分析（ChatBI）功能，简化数据提取和可视化过程。 Key features: - 实现对话式数据分析（ChatBI） - 快速提炼获取所需的数据信息及可视化图表 - 支持进一步智能分析 - 开箱即用，无需复杂开发 - 提供安全可控的工作空间资源隔离机制 - 支持细粒度数据权限配置 - 支持多种集成方式，快速嵌入到 n8n、Dify、MaxKB、DataEase 等应用 - 高效运营，基于用户交互数据持续迭代优化 Tech stack: JavaScript Why it matters: SQLBot 提供了一种简单且高效的方式来处理复杂的数据分析任务，通过结合大语言模型和 RAG 技术，大大提高了数据提取和分析的效率，同时确保了数据的安全性和合规性。

### Overview: Donut is an OCR-free document understanding transformer that achieves state-of-the-art performance across various visual document understanding tasks. It is based on the academic paper "OCR-free Document Understanding Transformer" from ECCV 2022, providing detailed experimental results and analyses. ### Key Features: - **OCR-free Document Understanding Transformer (Donut)**: A method that utilizes an OCR-free end-to-end Transformer model to perform document understanding tasks. - **Synthetic Document Generator (SynthDoG)**: A tool that helps the model pre-train to be flexible on various languages and domains. - **Official Implementation**: Provides full experimental results and analyses, including pre-trained models and web demos. ### Tech Stack: - Python ### Why It Matters: Donut and SynthDoG offer a new approach to document understanding that does not require off-the-shelf OCR engines, yet still achieves high performance. This method is particularly useful for applications where OCR is not feasible or desirable, such as in environments with limited access to OCR tools. ### Repository: - **Repository**: clovaai/donut - **Stars**: 6927 - **Description**: Official Implementation of OCR-free Document Understanding Transformer (Donut) and Synthetic Document Generator (SynthDoG), ECCV 2022 - **Primary Language**: Python - **Downloads**: 2.2k - **PyPI**: pip install donut-python - **Demo**: Gradio web demos available at [Gradio space web demo](https://huggingface.co/sp

Overview: This project introduces a PyTorch implementation of a method for high-resolution image-to-image translation, specifically focusing on synthesizing and manipulating 2048x1024 images using conditional Generative Adversarial Networks (GANs). Key features: - Synthesizing and manipulating 2048x1024 images with conditional GANs - Utilizing semantic label maps to generate photo-realistic images - Generating portraits from face label maps - Interactive editing capabilities Tech stack: The project is implemented using PyTorch. Why it matters: This method offers significant advancements in high-resolution image manipulation, making it particularly useful in applications such as creating photorealistic streetview images and generating realistic portraits from facial feature maps.

Overview: A.I.G (AI-Infra-Guard) is a comprehensive AI Red Teaming platform designed by Tencent Zhuque Lab, offering advanced security scanning and evaluation capabilities for AI ecosystems. Key features: - ClawScan (OpenClaw Security Scan) - Agent Scan - AI infra vulnerability scan - MCP Server & Agent Skills scan - Jailbreak Evaluation Tech stack: Python Why it matters: A.I.G is crucial for securing AI infrastructures, making it essential for developers and security professionals alike. Its features ensure robust protection against vulnerabilities and potential security breaches, making it a valuable tool in the AI security landscape.

### Overview: ClearML is a comprehensive suite for managing, orchestrating, and monitoring AI and machine learning workflows. It integrates experiment management, data management, pipeline orchestration, and serving capabilities. ### Key features: - **Experiment Manager**: Automates experiment tracking, environments, and results. - **MLOps / LLMOps**: Orchestration, automation, and pipelines for ML/DL/GenAI jobs. - **Data-Management**: Fully differentiable data management and version control on object-storage. - **Model-Serving**: Cloud-ready scalable model serving solution with GPU support and model monitoring. - **Reports**: Create and share rich Markdown documents for online content. - **Orchestration Dashboard**: Live dashboard for compute clusters. - **Fractional GPUs**: Container-based GPU memory limitation. ### Tech stack: Python ### Why it matters: ClearML simplifies the entire AI development and production process, from experimentation to model serving, making it easier and more efficient for developers and data scientists.

### Overview: AppAgent is an LLM-based multimodal agent framework designed to operate smartphone apps. It simplifies the interaction process by mimicking human-like actions like tapping and swiping, allowing it to navigate and use new apps without requiring backend system access. ### Key Features: - **Simplified Action Space**: Enables the agent to interact with apps through a simplified interface, similar to how humans operate smartphones. - **Autonomous Exploration**: Learns to navigate and use new apps through autonomous exploration or by observing human demonstrations. - **Knowledge Base**: Generates a knowledge base that the agent uses to execute complex tasks across different applications. - **GUI Agent with Evolving Mechanism**: The next-generation GUI Agent with an evolving mechanism. ### Tech Stack: Python ### Why It Matters: AppAgent simplifies app interaction, broadening its applicability across diverse apps without the need for backend system access. This makes it more accessible and user-friendly, enhancing the overall user experience.

### Overview: This repository is a curated collection of research papers and projects related to **OpenAI's o1** and reasoning techniques. It serves as a comprehensive resource for developers and researchers interested in the latest advancements in large language models and their reasoning capabilities. ### Key Features: - **Research Papers and Blogs**: Includes a wide range of papers and blogs from OpenAI, Anthropic, and other leading researchers. - **OpenAI o1 Mini**: Focuses on the o1 mini model, which is known for its cost-efficient reasoning capabilities. - **Reasoning Techniques**: Covers various reasoning techniques and approaches used in large language models, including reinforcement learning and self-supervised learning. ### Tech Stack: The repository is built using Markdown and GitHub, making it easy to read and navigate. ### Why It Matters: This collection is essential for developers and researchers who are interested in staying updated with the latest developments in large language models and their reasoning capabilities. It provides a valuable resource for understanding and implementing advanced reasoning techniques in AI systems.

### Overview: CodeCompanion.nvim is an AI-powered coding assistant designed for Neovim users. It integrates with various large language models and agents, including Anthropic, OpenAI, Gemini, and more, to enhance your coding experience. ### Key features: - **Copilot Chat**: Combines the power of Copilot with Zed AI, making it easier to write code. - **Neovim Integration**: Seamless integration with Neovim for seamless workflow. - **Agent Support**: Built-in support for LLMs from Anthropic, DeepSeek, Google Gemini, GitHub Copilot, and more. - **Agent Client Protocol (ACP)**: Enables coding with agents like Augment Code, Cagent, Claude Code, Codex, Copilot CLI, Gemini CLI, and more. - **Model Context Protocol (MCP)**: Supports agents like Claude Code, Codex, and others. - **Extensibility**: Supports bringing your own LLMs and agents, and integrating with Agent Client Protocol and Model Context Protocol. ### Tech stack: Lua ### Why it matters: CodeCompanion.nvim offers a powerful and versatile tool for developers, enhancing productivity and efficiency in coding workflows. Its integration with multiple LLMs and agents makes it a valuable addition to any developer's toolkit. ### Repository: olimorris/codecompanion.nvim ### GitHub stars: 6899 ### GitHub description: ✨ AI Coding, Vim Style ### Sponsors: Thank you to the following people who support CodeCompanion.nvim: - Alexander Garcia

Overview: Mooncake is a KVCache-centric architecture designed specifically for LLM serving, serving as the platform for Kimi, a leading LLM service provided by Moonshot AI. Key features: - Innovative architecture for handling real workloads - Enables Kimi to handle 75% more requests while adhering to SLOs - Supports disaggregated serving with Mooncake Store, TokenSpeed, and vLLM integration Tech stack: C++ Why it matters: Mooncake's architecture optimizes performance and scalability, making it essential for LLM serving in high-demand environments.

### Overview: 《大模型与智能体》电子书与6个循序渐进的编程任务，包括Transformer、mini-GPT、SFT/DPO、RAG、工具调用与编程智能体。本书通过6个任务逐步介绍大模型与智能体的构建过程，涵盖了原理、系统构建、应用与治理等内容。本书适合Python和深度学习基础的开发者阅读。 ### Key features: - **任务1: Transformer** - 手写注意力机制与Transformer，用于中文情感分类任务。 - **任务2: mini-GPT** - 训练小型语言模型，实现分词、RoPE、KV Cache与采样。 - **任务3: SFT/DPO** - 手写LoRA，完成SFT与DPO，对比微调效果。 - **任务4: RAG** - 构建检索、重排与生成流程，并评测问答质量。 - **任务5: 工具调用 Agent** - 实现ReAct循环、工具调用与错误恢复。 - **任务6: Mini Coding Agent** - 构建能修改代码、运行测试并迭代的编程智能体。 ### Tech stack: Python ### Why it matters: 通过6个循序渐进的编程任务，读者可以深入了解大模型与智能体的构建过程，从原理到工程实现，再到应用与治理。本书适合Python和深度学习基础的开发者阅读，帮助读者掌握构建智能体所需的技能和知识。 ### Repository: n

Overview: This project demonstrates the application of Deep Q-Networks to improve the performance of the classic game Flappy Bird. By using raw pixel inputs, the algorithm successfully learns to navigate the game without the need for a complex environment. Key features: - Flappy Bird hack using Deep Reinforcement Learning - Utilizes Deep Q-Networks for training - Demonstrates faster convergence with preprocessing techniques Tech stack: Python, TensorFlow, pygame, OpenCV-Python Why it matters: This project showcases the potential of Deep Q-Networks to solve complex tasks, such as games, by leveraging raw pixel inputs and efficient training methods.

Overview: PostgresML is a Postgres extension that integrates machine learning and AI functionalities directly within the database, leveraging GPU acceleration for faster computations and model inference. Key features: - In-Database ML/AI - GPU Acceleration - Large Language Models - RAG Pipeline - Vector Search - Diverse ML Algorithms - High Performance - Scalability - NLP Tasks - Security - Seamless Integration Tech stack: Rust Why it matters: PostgresML enables efficient, reliable, and scalable ML/AI operations directly within the database, eliminating the need for separate systems and data transfers, making it ideal for applications that require both data storage and machine learning inference. Repository: postgresml/postgresml GitHub stars: 6823 GitHub description: Postgres with GPUs for ML/AI apps.