AI Agents & Automation

RLSeq2Seq is an open-source project developed in TensorFlow, focusing on deep reinforcement learning (RL) for sequence-to-sequence (seq2seq) models. It addresses common problems in seq2seq models such as exposure bias and train/test inconsistency by integrating RL methods. The repository provides code for implementing various models, including Scheduled Sampling, Soft-Scheduled Sampling, End2EndBackProp, Policy-Gradient with Self-Critic learning, and Actor-Critic models using DDQN and Dueling networks. It is particularly geared towards abstractive text summarization, offering helper codes for processing datasets like CNN/Daily Mail and Newsroom. The project is suitable for researchers and developers looking to explore and apply advanced RL techniques to improve seq2seq model performance.

Enso provides a platform for businesses to either build their own AI agents or acquire pre-built ones to automate various aspects of their operations. The tool aims to facilitate the creation of an autonomous business by leveraging AI technology. While specific features are not detailed on the homepage, the core offering revolves around the deployment and management of AI agents to streamline workflows and enhance efficiency. Enso positions itself as a solution for businesses looking to integrate advanced AI capabilities without necessarily developing them from scratch, offering a pathway to digital transformation and operational autonomy.

RoutePerfect is an online trip-planning platform that leverages AI and human expertise to help travelers design and customize their perfect trips. Users can choose from three easy planning options, including an AI Trip Planner, to create itineraries tailored to their preferences, travel style, and budget. The platform allows for easy booking of flights, accommodations, transportation, and activities through leading travel partners, enabling users to compare prices and access loyalty program benefits. RoutePerfect also enriches trips with exclusive travel perks and offers a wide selection of popular itineraries for various destinations like Italy, Japan, and Germany. It aims to provide a seamless and optimized travel planning experience from start to finish.

RLCard is a comprehensive, open-source toolkit designed for reinforcement learning (RL) in card games. Developed by DATA Lab at Rice and Texas A&M University, it offers a versatile platform for researchers and developers to implement and test various RL and searching algorithms within popular card game environments such as Blackjack, Leduc Hold'em, Texas Hold'em, DouDizhu, Mahjong, UNO, Gin Rummy, and Bridge. The toolkit provides easy-to-use interfaces, supports environment local seeding, multiprocessing, and includes a model zoo with pre-trained and rule-based models. It also integrates with PettingZoo, allowing for multi-agent reinforcement learning experiments.

Mage-AI is an open-source platform designed for building, running, and managing data pipelines efficiently. It offers a self-hosted development environment that enables teams to create production-grade data pipelines using Python, SQL, or R in a modular, notebook-style UI. Key capabilities include automating ETL tasks, orchestrating data transformations, and connecting to various data sources like databases, APIs, and cloud storage with prebuilt connectors. The tool supports visual debugging with logs and step-by-step execution, and allows for manual or scheduled job execution. For advanced needs, Mage Pro offers enterprise orchestration, collaboration, AI-powered workflows, and robust features like multi-environment orchestration and real-time monitoring.

RL-Factory is an open-source framework designed for efficient reinforcement learning (RL) post-training in Agentic Learning. It significantly simplifies the process by decoupling the environment from RL post-training, allowing users to train agents with only a tool configuration and a reward function. A key differentiator is its support for asynchronous tool-calling, which makes RL post-training up to 2x faster than existing frameworks. The platform natively supports one-click DeepSearch training, multi-turn tool-calling, model judge reward mechanisms, and training for various models, including Qwen3. Future updates aim to introduce a WebUI for data processing, environment definition, and project management, alongside support for more models and multimodal agentic learning.

RagaAI Inc. delivers production-grade AI Agent Suites specifically designed for critical sectors like Healthcare, Lifesciences, and Aerospace. The platform is engineered to provide maximum reliability, leveraging its proprietary Prism and Catalyst technologies. These AI Agent Suites are purpose-built to address complex challenges, offering solutions for areas such as Radiology, Allergy, and Clinical Trials. RagaAI emphasizes proven ROI and enterprise-grade reliability, aiming to empower healthcare leaders and other industry professionals with advanced AI capabilities. The tool focuses on reducing operational inefficiencies, as evidenced by claims of a 46.5% reduction in claim denials and 3x faster radiology review.

CUGA Agent is a Configurable Generalist Agent developed by IBM Research, recognized as a leader in the AppWorld Benchmark. This tool allows users to define a task with a brief description, and the agent autonomously executes the necessary steps. It features the ability to connect to live enterprise demo applications, perform requested actions, and return results. Designed to facilitate faster deployment of AI solutions, CUGA Agent provides a foundation for generalist AI that can be customized for specific domains, making it a versatile solution for various operational needs.

real_time_face_recognition is an open-source project designed for real-time face detection and recognition, built upon a combination of powerful libraries including OpenCV, TensorFlow, MTCNN, and Facenet. The system utilizes MTCNN for efficient face detection and Facenet for generating face embeddings, enabling robust recognition capabilities. While offering advanced features for computer vision tasks, it's important to note that this repository is no longer actively maintained. Users can explore its functionalities through Jupyter Notebook examples, requiring dependencies like TensorFlow and OpenCV with Python bindings.

ad_examples is a comprehensive, open-source collection of anomaly detection methods, encompassing iid (point-based), graph, and time series data. The repository features advanced techniques such as active learning for anomaly detection/discovery, Bayesian rule-mining, and methods for diversity, explanation, and interpretability. It also includes analysis of incorporating label feedback with ensemble and tree-based detectors, and explores adversarial attacks using Graph Convolutional Networks. The tool supports various standard unsupervised anomaly detectors like Isolation Forest, LODA, One-class SVM, and LOF, alongside forecasting-based methods for time series data. It's designed for researchers and developers working with anomaly detection, offering both theoretical insights and practical examples.

Unbody Lab is dedicated to questioning, exploring, experimenting, and building adaptive thinking tools. It challenges the traditional software paradigm where products are rigid and users adapt to them. Unbody envisions AI as a means to unlock latent human potential, surfacing what's already present but hard to access, such as patterns, blind spots, and capacity to act. The platform aims to create a cognitive exoskeleton that extends memory, sharpens attention, and fosters clarity. It prioritizes returning time to users rather than capturing it, and builds tools that adapt to the human, learning rhythms and respecting limits. Unbody also emphasizes calibrated friction, ensuring tools support intentions without competing for attention, and values craft as care, absorbing complexity so users don't have to.

Agents.jl is a pure Julia general-purpose framework designed for agent-based modeling (ABM), a computational simulation methodology where autonomous agents interact with their environment and other agents based on predefined rules. Key highlights include its simplicity, extensive interface with thousands of possible agent actions, and high performance, often surpassing established competitors. The framework supports simulations on Open Street Maps and allows for both traditional discrete-time ABM simulations and continuous-time event queue-based ABM simulations. It also provides native integration with Reinforcement Learning (RL) for advanced applications. Agents.jl is part of JuliaDynamics, an organization focused on creating high-quality scientific software.

Cimba.ai is an AI-native agentic command center designed for enterprise finance and business operations. It operationalizes intelligence by combining trusted data, business context, and structured workflows into a single operational system. The platform enables teams to create governed AI agents and repeatable workflows that actively analyze data, answer questions, and deliver proactive, trusted next best actions. Key features include audit logs and traceability for SOX/SOC 2 compliance, cross-signal insights, and enterprise data integrations. Cimba helps finance teams with forecasting and variance analysis, customer success teams with health and retention monitoring, and operations teams with performance monitoring and anomaly investigation, scaling insights without endless dashboards.

JTP 3 Hydra Demo is a Hugging Face Space by RedRocket designed to analyze images. Users can upload an image or provide a URL, and the application will process it to identify and highlight key features. The tool offers adjustable sensitivity and depth of analysis, allowing users to customize the results to their specific needs. This makes it suitable for detailed image examination and feature extraction. It operates within the Hugging Face ecosystem, leveraging its infrastructure for deployment and compute resources.

The Gradient is an AI-native product design agency specializing in building UX/UI for AI products across various sectors including fintech, healthcare, and retail. Unlike traditional design firms, they deliver working prototypes built with real models, real data, and real behavior, not just mockups. Their services include AI Transformation, helping companies explore where AI fits and rebuild products around it, and UI for AI, assisting with new AI product development or improving existing ones. With nearly a decade of experience in designing products used by millions globally, The Gradient brings deep user and domain understanding to AI-native design, ensuring human intention connects seamlessly with machine capability.

Destinus, previously known as Daedalean AI, is a European defense manufacturer focused on innovating autonomous flight systems, advanced propulsion, and AI-driven software. The company designs and produces strike and air defense systems, including cruise missiles, loitering munitions, and kinetic interceptors, specifically for European and allied forces. Their core strength lies in combining a unified autonomy stack, deep vertical integration, and industrial-scale manufacturing to rapidly develop and deploy capabilities. Destinus aims to strengthen Europe’s security and strategic autonomy by delivering advanced, affordable systems to protect people and critical infrastructure, operating with a mission of "Defence at Industrial Tempo" and a vision for a secure Europe built on autonomy, scale, and industrial strength.

XBOW is an autonomous offensive security platform designed to transform penetration testing into a machine-scale offensive security system. It executes targeted attacks autonomously, allowing security teams to explore deeper attack paths than traditional pentesting. Every potential finding is independently validated through real exploitation, providing clear, reproducible proof without manual validation cycles. This frees security teams to focus on investigation, judgment, and remediation. XBOW aims to reduce real breach risk, shorten the path from test to fix, keep pace with modern development, and meet compliance with confidence. The platform offers on-demand and continuous penetration testing plans, including comprehensive compliance-ready reports for various frameworks.

openmlsys-zh is the Chinese version of the open-source textbook "Machine Learning Systems: Design and Implementation." This comprehensive resource delves into the core principles and practical implementation experiences of modern machine learning systems. It covers a wide array of topics, including programming interfaces, computation graphs, compilers, and distributed training. The book is designed to be valuable for students looking to deepen their understanding of ML systems, researchers developing custom operators or large models, and engineers responsible for ML infrastructure and performance tuning. The project is hosted on GitHub and provides detailed guidance for building and contributing to the content.

RFBNet is an open-source Receptive Field Block Net designed for accurate and fast object detection, presented at ECCV 2018. Inspired by human visual systems, it proposes a novel RFB module that considers the relationship between the size and eccentricity of receptive fields to improve feature discriminability and robustness. The RFB module is integrated into SSD with a lightweight CNN model, forming the RFB Net detector. The repository provides code for training and evaluating RFB Net for object detection, supporting datasets like VOC and COCO. It offers various models including RFBNet300, RFBNet512, and RFB MobileNet, with pre-trained weights available for download. The tool is implemented in PyTorch and supports Python 3+.

SMOKE is an open-source, real-time monocular 3D object detector specifically designed for autonomous driving applications. It utilizes a single-stage approach via keypoint estimation to achieve efficient and accurate 3D object detection from a single camera. The system boasts a runtime of approximately 30ms on an NVIDIA TITAN XP GPU. SMOKE's performance has been evaluated on the KITTI 3D detection dataset, showing competitive results for detecting cars, pedestrians, and cyclists. The project provides pretrained weights and detailed setup instructions for researchers and developers, building upon components from CenterNet, maskrcnn-benchmark, and Detectron2.

TweetDetective is a powerful Chrome extension designed to identify AI-generated content on Twitter. Utilizing advanced algorithms, it analyzes tweets and provides a real-time probability percentage, indicating the likelihood of AI text. This tool boasts a 97% accuracy rate for longer texts and aims to help users maintain authenticity and distinguish between genuine human interactions and automated postings. It integrates seamlessly into the Twitter interface, allowing users to see AI detection results directly on tweets. TweetDetective supports English content and offers a simple installation and API key setup process for immediate use.

Steel is an open-source, embeddable Scheme interpreter implemented in Rust, designed for scripting and extending Rust applications. It functions as a bytecode virtual machine and is largely compliant with the R5RS standard, with R7RS support currently under development. Key features include support for modules using `require` and `provide` (similar to Racket), `syntax-rules` and `syntax-case` macros, and seamless integration with native Rust functions and structs via embedding or FFI. It also provides higher-order contracts and built-in immutable data structures like lists, vectors, hashmaps, and hashsets. A standalone interpreter/REPL is included, and it can be installed via Cargo, Nix, or used through an online playground. The API is considered relatively stable but may undergo changes before version 1.0.

AgentVerse is a comprehensive platform designed for the lifecycle management of AI agents. It enables users to launch and manage both native and external agents, facilitating their integration with ASI:One. The platform offers robust features for agent discovery, allowing users to configure metadata and indexing settings to enhance visibility across ASI:One search. Furthermore, AgentVerse provides tools for agent optimization, where users can analyze queries, logs, and responses to debug behavior and improve agent accuracy. It also includes performance tracking capabilities to monitor query volume and ranking visibility over time. The platform aims to simplify the process of deploying and maintaining AI agents, ensuring they perform effectively and are easily discoverable.

TaxiDisponible is a pioneering French platform designed to connect passengers directly with verified taxi drivers across more than 15 major cities in France. Unlike traditional VTC applications that charge high commissions, TaxiDisponible operates on a unique 0% commission model, allowing drivers to retain 100% of their earnings. Passengers can easily search for taxis by city, browse detailed driver profiles including reviews and vehicle types, and contact their chosen driver directly via phone or WhatsApp. This eliminates the need for a separate application download and ensures transparent, state-regulated fares without surge pricing. The service is available 24/7 and caters to various needs, including airport transfers, train station pickups, business travel, medical transport, long-distance journeys, and even parcel delivery.