Projects

### Overview: This repository is a curated list of public-source, research, and commercial tools designed for AI security and AI-assisted cybersecurity. It includes tools for autotriage of security findings, AI agent and coding-agent security, AI/ML supply chain and model security, pentest and red-team agents, AI-powered recon and narrow ML tools, AI-powered SAST and secure code review, LLM-driven fuzzing, threat intelligence, log analysis, reverse engineering, LLM red-teaming, and more. ### Key Features: - **Metis:** A security-review framework that triages external SARIF findings using source-navigation. - **AI Agent Security:** Tools for scanning and auditing security, frameworks and rule standards, runtime protection and enforcement. - **AI/ML Supply Chain & Model Security:** Tools for securing AI/ML supply chains and models. - **Pentest & Red-Team Agents:** Agents for penetration testing and red-teaming. - **AI-Powered Recon & Narrow ML Tools:** - **Subdomain & DNS Prediction:** Tools for predicting subdomains and DNS predictions. - **Recon Screenshot Triage:** Tools for triaging screenshots from recon tools. - **Software / Tech Fingerprinting:** Tools for identifying software and technology fingerprints. - **AI-Assisted Fuzzing:** Tools for AI-assisted fuzzing. - **Password / Credential ML:** Tools for password and credential ML. - **Phishing Detection (Visual / URL):** Tools for detecting phishing URLs and visual phishing. - **AI/ML-Ass

Overview: PyHealth is a comprehensive deep learning toolkit designed for healthcare applications, offering a user-friendly interface and robust features for model development, training, and deployment. Key features: - Model development and training - Model deployment and inference - Data preprocessing and augmentation - Model evaluation and performance metrics - Model sharing and collaboration Tech stack: Python Why it matters: PyHealth simplifies the process of developing, training, and deploying deep learning models in healthcare applications, making it an essential tool for researchers, developers, and practitioners in the field.

Overview: Tika-Python is a Python binding to the Apache Tika REST services, enabling seamless integration of Tika capabilities into the Python ecosystem. Key features: - Native Python bindings for Apache Tika - Supports various Tika REST services - Easy installation via pip - Background Tika server starts on installation Tech stack: Python Why it matters: Tika-Python simplifies the integration of Tika's powerful text analysis and extraction capabilities into the Python programming language, making it easier for developers to leverage Tika's features without needing to install and manage the Tika server separately.

### Overview: MMaDA is a cutting-edge multimodal diffusion language model designed to excel in various domains, including textual reasoning, multimodal understanding, and text-to-image generation. ### Key Features: - **Unified Diffusion Architecture**: Eliminates modality-specific components, achieving superior performance across diverse domains. - **Mixed Long CoT Fine-Tuning**: Curates a unified CoT format across modalities, enhancing coherence and consistency. - **UniGRPO Policy-Gradient RL Algorithm**: Tailored for diffusion foundation models, ensuring consistent performance improvements. ### Tech Stack: Python ### Why it Matters: MMaDA offers a unified approach to diffusion models, improving performance across various tasks and domains, making it a significant advancement in the field of language modeling and multimodal AI.

Overview: TorchQuantum is a PyTorch-based framework designed for simulating quantum classical computations. It supports statevector and pulse simulation on GPUs, enabling scalable quantum computation on classical hardware. Key features: - Easy construction and simulation of quantum circuits in PyTorch - Dynamic computation graph for easy debugging - Gradient support via autograd - Batch mode inference and training on CPU/GPU - Easy deployment on real quantum devices like IBMQ - Easy hybrid classical-quantum model construction - (coming soon) Pulse-level simulation Tech stack: PyTorch Why it matters: TorchQuantum simplifies the process of quantum classical simulation and quantum machine learning, making it accessible for researchers and developers interested in quantum computing. Repository: mit-han-lab/torchquantum Primary language: Jupyter Notebook GitHub stars: 1674 GitHub description: A PyTorch-based framework for Quantum Classical Simulation, Quantum Machine Learning, Quantum Neural Networks, Parameterized Quantum Circuits with support for easy deployments on real quantum computers. README excerpt: <p align="center"> <img src="torchquantum_logo.jpg" alt="torchquantum Logo" width="450"> </p> <h2><p align="center">Quantum Computing in PyTorch</p></h2> <h3><p align="center">Faster, Scalable, Easy Debugging, Easy Deployment on Real Machine</p></h3> <p align="center"> <a href="https://torchquantum.readthedocs.io/">

Overview: jiant is an NLP toolkit designed for multitask and transfer learning research in natural language processing. Key features: - Supports multitask learning - Supports transfer learning - Supports 50+ natural language understanding tasks - Supports benchmarks including GLUE, SuperGLUE, and XTREME Tech stack: Python Why it matters: jiant provides a comprehensive toolkit for researchers in the field of natural language processing, enabling them to easily incorporate multitask and transfer learning techniques into their projects. This makes it an essential tool for advancing research in NLP.

Overview: Magnitude is a fast, efficient, and simple Python package designed for utilizing vector embeddings in machine learning models. It offers unique features such as out-of-vocabulary lookups and streaming of large models over HTTP. Key features: - Fast and efficient vector embedding utility - Simple and easy-to-use interface - Supports out-of-vocabulary keys - Handles misspellings and typos - Provides streaming of large models over HTTP Tech stack: Python Why it matters: Magnitude is a faster and simpler alternative to Gensim, making it ideal for non-NLP applications. Its unique features, such as out-of-vocabulary lookups and streaming, make it highly valuable for various domains beyond NLP. Repository: plasticityai/magnitude Primary language: Python GitHub stars: 1674 GitHub description: A fast, efficient universal vector embedding utility package. README excerpt: A feature-packed Python package and vector storage file format for utilizing vector embeddings in machine learning models in a fast, efficient, and simple manner developed by [Plasticity](https://www.plasticity.ai/).

### Overview: This project documents the process of building a real-time hand detector using neural networks and Tensorflow Object Detection API. The goal is to demonstrate how neural networks can be applied to tracking hands in various views, including egocentric and other perspectives. The project includes scripts and steps for training a hand detector on a dataset, and provides code that can be adapted for other use cases. ### Key Features: - Tensorflow Object Detection API - Real-time hand detection on video streams - Adaptability to other use cases - Code for exporting frozen graph for model compatibility with different versions of Tensorflow ### Tech Stack: - Python - Tensorflow ### Why It Matters: This project provides a practical demonstration of how neural networks can be applied to the challenging problem of tracking hands in various views. The code and scripts included in the repository can be adapted for other use cases, making it a valuable resource for developers looking to integrate hand tracking into their projects. ### Repository: victordibia/handtracking ### Primary Language: Python ### GitHub Stars: 1674 ### GitHub Description: Building a Real-time Hand-Detector using Neural Networks (SSD) on Tensorflow ### README Excerpt: This repo documents steps and scripts used to train a hand detector using Tensorflow (Object Detection API). The goal is to demonstrate how neural networks can be applied to the (hard) problem of tracking hands (egocentric and other views). The code and scripts included in the repository can be adapted for other use cases, making it a valuable resource for

Overview: Rainbow is a deep reinforcement learning algorithm that combines several improvements to enhance its performance in deep reinforcement learning tasks. It is designed to be data-efficient and capable of learning from limited data. Key features: - DQN - Double DQN - Prioritised Experience Replay - Dueling Network Architecture - Multi-step Returns - Distributional RL - Noisy Nets Tech stack: Python Why it matters: Rainbow is significant because it addresses common limitations of deep reinforcement learning algorithms and provides a more efficient way to learn from limited data, making it particularly useful in scenarios where data is scarce or expensive to obtain.

Overview: Testkube is an open-source platform designed for AI-driven engineering teams, enabling them to run and analyze automated tests using their existing tools and Kubernetes infrastructure. Key features: - Execute any tests, including API, E2E, Performance, Security, and Infrastructure tests. - Trigger tests manually, on schedules, through CI/CD/GitOps pipelines, Kubernetes Events, and via the REST API. - Aggregate test results, artifacts, logs, and resource metrics for centralized troubleshooting and reporting. Tech stack: Go Why it matters: Testkube helps engineering teams streamline their testing processes by providing a unified platform for executing and analyzing various types of tests, making it easier to manage and troubleshoot issues across different environments.

Overview: Frustum PointNets is a novel 3D object detection method that leverages 2D object detectors and PointNet architectures to achieve high accuracy in RGB-D data. This project was presented at CVPR 2018 and significantly outperformed previous state-of-the-art methods on KITTI and SUNRGBD benchmarks. Key features: - Combines 2D object detectors with state-of-the-art 3D deep learning techniques. - Uses PointNet/PointNet++ networks to directly work on 3D point clouds. - Proposes a pipeline that reduces 3D search space and leverages high-resolution image information. - Achieves high recalls for smaller objects like pedestrians and cyclists. Tech stack: Python Why it matters: Frustum PointNets is a significant advancement in 3D object detection, offering improved accuracy and efficiency compared to previous methods, especially in challenging scenarios like RGB-D data.

### Overview: ROSA, the AI Agent designed for interacting with ROS1- and ROS2-based robotics systems, helps developers inspect, diagnose, understand, and operate robots more efficiently. ### Key features: - **Natural Language Queries**: ROSA can interpret and respond to natural language commands to interact with robots. - **ROS Version Support**: ROSA is compatible with both ROS1 (Noetic) and ROS2 (Humble, Iron, Jazzy). - **Model Configuration**: The AI is adaptable and can be customized with different language models for better performance. ### Tech stack: Python ### Why it matters: ROSA simplifies robotics development by making it more accessible and efficient, allowing developers to focus on the core functionality of their robots rather than the complexities of ROS integration.

Overview: This repository contains a collection of n8n workflows, templates, and AI automations created and maintained by The Recap AI. Join our free AI Automation Mastery community to connect with AI builders and learn how to build. Key features: - Marketing Team Voice Agent - Dental Practice Voice Agent - WhatsApp AI Chatbot Agent - AI Gmail Agent - Auto Repair Shop Gmail Agent Tech stack: n8n Why it matters: Join our AI Automation Mastery community to connect with AI builders and learn how to build AI automations and workflows.

### Overview: This repository is a curated list of 180+ YouTube channels across various domains such as Data Science, Machine Learning, Artificial Intelligence, Computer Science, programming, software engineering, and more. It serves as a resource for individuals interested in learning and staying updated with the latest in these fields. ### Key Features: - **Data Science / Data Analysis:** Channels like "Ken Jee," "Tina Huang," and "Data Professor" offer valuable insights and tutorials. - **ML/AI/DL:** "Krish Naik," "Import Data," and "Chris Lovejoy" provide in-depth explanations and practical applications. - **Data Engineering:** "Imaad Mohamed Khan" and "Part Time Larry" offer practical tips and real-world examples. - **Statistics/Math:** "Alex The Analyst" and "Data Science Jay" provide statistical and mathematical analysis. - **AI Research:** "RichardOnData" and "The High ROI Data Scientist" focus on cutting-edge research and applications. - **Computer Science / Algorithms:** "Luke Barousse" and "Madhav Thaker" offer insights into computer science and algorithms. - **Programming:** "James Briggs" and "Chris Lovejoy" provide tutorials and practical examples. - **Web-dev (HTML, CSS, JS):" "Data Science Jay" and "Luke Barousse" offer web development tips. - **Software Engineering:** "Nate at StrataScratch" and "Data School" provide software engineering insights. - **Langauges:** "Python" and "R" channels offer tutorials and projects

Overview: Hopenet is an accurate and easy-to-use head pose estimation network trained on the 300W-LP dataset and tested on real data with good qualitative performance. Key features: - Accurate head pose estimation - Easy to use implementation - Models trained on 300W-LP dataset - Tested on real data with good qualitative results Tech stack: Python Why it matters: Hopenet provides a reliable and efficient method for estimating head pose in videos, making it useful for various applications such as video analysis, surveillance, and more.

### Overview: OpenChoreo is an open-source developer platform designed for Kubernetes, providing a comprehensive set of tools and abstractions for building, deploying, and operating applications. It includes development and platform abstractions, a Backstage-powered developer portal, CI/CD, GitOps, and observability features. ### Key Features: - **Development Abstractions**: Provides high-level abstractions for Kubernetes resources. - **Backstage Developer Portal**: A modern, user-friendly portal for developers. - **CI/CD**: Integrated Continuous Integration/Continuous Deployment pipelines. - **GitOps**: Git-based workflow management for Kubernetes. - **Observability**: Real-time monitoring and logging capabilities. - **Security**: Built-in security features and access controls. ### Tech Stack: Go ### Why It Matters: OpenChoreo simplifies the process of building, deploying, and operating applications on Kubernetes by providing a unified platform for developers and platform engineers. It addresses the challenges of maintaining and evolving a bespoke system, making it easier and more cost-effective for both developers and platform engineers.

Overview: Trench is an open-source analytics infrastructure built on ClickHouse, Kafka, and Node.js, designed for real-time event tracking and analytics. It is GDPR, PECR, and no-cookie compliant, making it ideal for building product analytics dashboards, LLM RAGs, observability platforms, or any other analytics product. Key features: - 🤝 Compliant with the Segment API (Track, Group, Identify) - 🐳 Deploy quickly with a single production-ready Docker image - 💻 Process thousands of events per second on a single node - ⚡ Query data in real-time - 🔗 Connect data to other destinations with webhooks - 👥 Open-source and MIT Licensed Tech stack: Trench is built on TypeScript, ClickHouse, Kafka, and Node.js. Why it matters: Trench is designed to handle large event volumes and provides real-time analytics, making it an ideal solution for building real-time product analytics dashboards, LLM RAGs, observability platforms, or any other analytics product. Its open-source nature and MIT license make it accessible and customizable for developers. Repository: FrigadeHQ/trench Primary language: TypeScript GitHub stars: 1665 GitHub description: Trench — Open-Source Analytics Infrastructure. A single production-ready Docker image built on ClickHouse, Kafka, and Node.js for tracking events. Easily build product analytics dashboards, LLM RAGs, observability platforms, or any other analytics product.

### Overview: This repository is a collection of Google Colab notebooks designed for quick and easy experimentation in machine learning (ML) projects. It provides a variety of notebooks across different ML models and techniques, making it an excellent resource for developers and researchers. ### Key Features: - **Google Colab Notebooks**: Includes notebooks optimized for Google Colab, ensuring seamless execution on Google's cloud platform. - **ML Models**: Covers a wide range of ML models such as transformers, image classification, and more. - **Techniques**: Features various ML techniques including transfer learning, sequence modeling, and more. ### Tech Stack: Python ### Why It Matters: This collection is highly useful for developers and researchers looking to experiment with ML models and techniques quickly and easily. The notebooks are optimized for Google Colab, ensuring efficient execution and a smooth development workflow.

### Overview: 齐伋體是一款源自明代木刻印刷书籍的 Ming typeface，由一名名为令东齐伋的16世纪木刻印刷者所命名。该项目旨在通过使用半自动计算机视觉和OCR技术，从木刻印刷书籍中提取并开发出这一珍贵字体。齐伋體旨在与古典中文编程语言wenyan-lang一起使用，是一个仍在开发中的开放源代码项目。 ### Key features: - 齐伋體是一款从明代木刻印刷书籍中提取并开发的Ming typeface。 - 使用半自动计算机视觉和OCR技术。 - 开源项目。 - 旨在与古典中文编程语言wenyan-lang一起使用。 - 周期性更新和开发。 ### Tech stack: Python ### Why it matters: 齐伋體作为一款珍贵的Ming typeface，从明代木刻印刷书籍中提取并开发，旨在与古典中文编程语言wenyan-lang一起使用。该项目是一个仍在开发中的开源项目，旨在通过使用半自动计算机视觉和OCR技术从木刻印刷书籍中提取并开发这一珍贵字体。

### Overview: This repository contains material related with architectures and models for synthetic data generation, focusing on Generative Adversarial Networks (GANs) and Gaussian Mixtures. It includes a comprehensive ecosystem for synthetic data generation, featuring various models for structure and time-series data. The source code is hosted on GitHub and can be installed via pip. ### Key features: - A new streamlit app for synthetic data generation with a user-friendly interface. - A fast synthetic data generation model based on Gaussian Mixture. - A conditional architecture for tabular data, specifically CTGAN, to simplify and improve the quality of synthetic data generation. ### Tech stack: TensorFlow 2.0 ### Why it matters: Synthetic data is artificially generated data that replicates the statistical components of real data, ensuring privacy and facilitating various applications such as privacy compliance, removing bias, balancing datasets, and augmenting datasets. It is essential for ensuring the quality and fairness of machine learning models, especially in scenarios where real data is not available or when privacy concerns are a concern.