Coding & Development

FederatedScope is a comprehensive and easy-to-use federated learning platform designed for both academic research and industrial applications. It offers flexible customization for a wide range of federated learning tasks, integrating rich functionalities through an event-driven architecture. The platform supports various applications including federated learning in computer vision (CV), natural language processing (NLP), and graph federated learning (GFL). It provides a DataZoo for automatic dataset download and preprocessing, and a ModelZoo with implementations of widely adopted model architectures. Users can run tasks in both standalone and distributed modes, with options for customized datasets and models.

FLUX LoRA DLC2 is a comprehensive collection of over 100 impressive LoRA models specifically designed for Flux. This tool, hosted on Hugging Face, allows users to easily generate images by typing a description and optionally selecting a LoRA style. It simplifies the process of customizing and fine-tuning AI models for various creative and AI-related tasks. The platform is accessible via a web interface, making it convenient for users to explore and utilize these diverse models to produce pictures that align with their textual prompts and chosen aesthetic.

The Indic NLP Library is a Python-based open-source project designed to provide resources and tools for Natural Language Processing (NLP) in various Indian languages. Recognizing the shared linguistic characteristics among Indian languages, this library offers a unified solution for common text processing needs. Key functionalities include text normalization, script information extraction, word tokenization and detokenization, sentence splitting, word segmentation, syllabification, script conversion, and Romanization/Indicization. It is built for Python 3.x and requires external data resources from the Indic NLP Resources project. The library is actively maintained and provides both a Python API and a command-line interface for ease of use, making it a valuable asset for developers and researchers working with Indian language data.

The Human & GPT-4 Evaluation of LLMs Leaderboard is a Hugging Face Space designed to benchmark and compare the performance of various Large Language Models (LLMs). It leverages both human and GPT-4 evaluations to provide a comprehensive assessment of AI model quality and capabilities. This open-source tool is valuable for researchers, developers, and data scientists who need to understand the strengths and weaknesses of different LLMs. While the live website currently displays a runtime error, the intent of the project is to offer a transparent and accessible platform for tracking advancements in LLM technology, aiding in the selection and development of more effective AI solutions.

Kaz LLM Leaderboard is a platform designed for evaluating Large Language Models (LLMs) specifically tailored for the Kazakh language. Users can submit benchmark results for various Kazakh language models in a standardized JSON format. The application then compiles these results into a comprehensive leaderboard, enabling researchers and developers to compare the performance of different LLMs. This tool is valuable for tracking advancements in Kazakh LLM development and identifying top-performing models for specific tasks. It serves as a centralized hub for the Kazakh AI community to share and assess their work.

La Leaderboard is an open-source tool hosted on Hugging Face Spaces, designed for evaluating and comparing the performance of large language models (LLMs). Its primary focus is on models that support languages spoken in Latin America and Spain, providing a crucial resource for researchers and developers working with these linguistic contexts. Users can access detailed leaderboard tables to view and compare model scores across different tasks, facilitating informed decisions on model selection and development. The platform aims to foster transparency and progress in multilingual LLM research by offering a standardized evaluation framework.

LLM Guard Playground is an AI security tool hosted on Hugging Face Spaces, developed by Protect AI. It provides an interactive environment for users to explore and assess the security of large language models. The platform allows for viewing and interacting with data applications, enabling users to experiment with various security measures and identify potential vulnerabilities within LLMs. While the current live website indicates a build error, the tool's intent is to offer a playground for security research and development related to AI models. It is available for free, making it accessible for developers and researchers interested in LLM security.

SingularityNET is a pioneering platform dedicated to building a decentralized AI infrastructure, fostering the development of Artificial General Intelligence (AGI) and Artificial Superintelligence (ASI). It offers several key components, including ASI:Chain, an AI-native Layer-1 blockchain for secure interoperability and scalable consensus; ASI:Create, a launchpad for developers to build, fund, and deploy AI agents; and ASI:Cloud, providing AI and GPU cloud services for builders and enterprises. The platform also supports open research initiatives for beneficial AGI/ASI and is a founding member of the Artificial Superintelligence Alliance. SingularityNET utilizes the FET (ASI) token to power its ecosystem, enabling access to AI services, governance participation, and network security.

MADRL is a repository offering code for multi-agent deep reinforcement learning (MADRL), providing implementations of several multi-agent reinforcement learning environments. These include Pursuit Evasion, Waterworld, Multi-Agent Walker, and Multi-Ant. The package requires OpenAI Gym and a forked version of rllab (the multiagent branch) for its functionality. It is designed for researchers and developers in the field of multi-agent reinforcement learning, allowing them to set up and run simulations with curriculum learning. The project also provides details on policy definitions and includes a citation for its accompanied paper, making it a valuable resource for academic and practical applications in MADRL.

LLM Gateway is an open-source API gateway designed to streamline the management and analysis of Large Language Model (LLM) requests. It acts as a middleware between applications and various LLM providers, including OpenAI, Anthropic, and Google Vertex AI. Key functionalities include routing requests to different providers, centralizing API key management, and tracking token usage and costs. The platform also provides performance monitoring and usage analytics to help users optimize their LLM interactions. It offers a unified API interface compatible with the OpenAI API format for seamless integration and supports both hosted and self-hosted deployment options.

MaxText is a high-performance, highly scalable, open-source library for Large Language Models (LLMs), implemented in pure Python/JAX. It is designed to run efficiently on Google Cloud TPUs and GPUs, supporting both pre-training and scalable post-training with techniques like Supervised Fine-Tuning (SFT) and Reinforcement Learning (GRPO, GSPO). MaxText achieves high Model FLOPs Utilization (MFU) and tokens/second across various cluster sizes, leveraging the power of JAX and the XLA compiler. It offers a library of high-performance models including Gemma, Llama, DeepSeek, Qwen, and Mistral, and serves as a launching point for ambitious LLM projects in research and production. Users can experiment with MaxText out of the box or fork and modify it to meet specific needs, with support for multi-modal training.

mergekit is a robust toolkit designed for merging pre-trained large language models, offering an out-of-core approach that enables complex merges even in resource-constrained environments. It supports execution entirely on CPU or with acceleration using as little as 8 GB of VRAM. The toolkit features a wide array of merging algorithms, including Linear, SLERP, TIES, and DARE, with more methods continuously being added. Key capabilities include support for various model architectures like Llama and Mistral, lazy loading of tensors for low memory use, and advanced techniques like piecewise assembly of models and Mixture of Experts merging. It also facilitates multi-stage merging and raw PyTorch model merging, making it a versatile solution for AI researchers and developers looking to combine model strengths without extensive retraining.

FinPlusTech, Inc. is an innovation-driven fintech company focused on providing cutting-edge AI technology and governance solutions for financial institutions. They specialize in developing enterprise AI systems, covering everything from initial planning and model development to ML operations and robust AI governance frameworks. Their flagship product, riskAI, is an AI-powered risk management system designed for banks, funds, and other financial entities to identify and manage tail-end risks and potential black swan events. FinPlusTech aims to help financial institutions rapidly develop and govern their AI programs, ensuring reliability, compliance, and responsibility in their AI deployments.

MAmmoTH2 is presented as the strongest open-source large language model (LLM) specifically designed for reasoning tasks. Hosted on Hugging Face Spaces by TIGER-Lab, this tool allows users to interact with a chatbot that generates text responses based on conversation history and a provided system prompt. It is primarily intended for research and development, enabling users to explore advanced language understanding and complex problem-solving. As an open-source model, MAmmoTH2 offers flexibility for developers and researchers to integrate and adapt it for various applications.

ocrbase is a lightweight, model-agnostic OCR API designed for standardizing document parsing across various visual language models (VLMs). It allows users to convert PDF and image documents into structured data formats such as Markdown and JSON. The tool is self-hostable and integrates with leading OCR models like PaddleOCR and GLMOCR, which boast high accuracy on benchmarks. Key features include API endpoints for parsing documents, asynchronous job processing, and structured data extraction. It also offers optional S3 input staging and BullMQ parse queuing for enhanced scalability and reliability, making it suitable for developers needing robust document processing capabilities.

Waylens Inc. is dedicated to improving road safety within the transportation technology sector by leveraging advanced AI video technology. The company equips video telematics providers with comprehensive tools designed to enhance driver safety and optimize fleet management. Waylens offers a suite of products including sophisticated AI models, specialized hardware, integrated fleet systems, efficient data workflows, and developer kits. These offerings enable businesses to implement cutting-edge solutions for monitoring and managing their fleets, ultimately contributing to safer transportation operations and more efficient logistics.

ProtTrans offers state-of-the-art pre-trained language models specifically designed for proteins. These models were developed using extensive computational resources, including thousands of GPUs from Summit and hundreds of Google TPUs, leveraging various Transformer architectures. The tool is invaluable for researchers in bioinformatics and drug discovery, providing functionalities such as feature extraction, fine-tuning of models for specific tasks, and making predictions related to protein structure and function. ProtTrans aims to support the bioinformatics community, including contributions to COVID-19 research, by regularly updating its repository with new pre-trained models and resources.

Nextpy is an open-source framework designed for building self-modifying AI software, currently in a 'just for friends' development stage. It emphasizes guardrails to define AI system boundaries and offers a powerful prompt engine for greater control over Large Language Models (LLMs) compared to traditional methods. Key capabilities include pre-compiling prompts, maintaining session state with LLMs for efficiency, and optimizing token usage. The framework is built for better AI generations, especially optimized for code generation, and can detect and fix syntax errors in LLM-generated code. It boasts modularity, multiplatform support, and a developer-first approach, aiming for 4-10x faster performance than alternatives like Streamlit.

Open CoT Leaderboard is a platform designed to track, rank, and evaluate the Chain-of-Thought (CoT) quality of open large language models (LLMs). Hosted as a Hugging Face Space, it provides a centralized location for researchers and developers to browse and filter a leaderboard of LLM benchmarks. Users can submit their own models for evaluation, allowing for comparison against existing models and contributing to the collective understanding of LLM performance. The platform offers transparency into the evaluation process and the status of submitted models, making it a valuable resource for identifying top-performing open-source LLMs and advancing AI research.

RoboHouse serves as a TU Delft fieldlab dedicated to advancing robotics and AI. It provides a collaborative environment where companies, scientists, and startups can develop, test, and refine new robotic and AI technologies. The initiative focuses on fostering innovation for a better society, particularly in areas where robots interact beneficially with humans. RoboHouse supports various activities including research, development, and project leadership, aiming to create real-world impact through its programs and facilities. It also offers opportunities for careers and partnerships within the robotics and AI sectors.

MININGLAMP Technology is a leading provider of enterprise data intelligence application software in China, with 19 years of experience in enterprise services. The platform focuses on extracting data value to help organizations achieve intelligent transformation and commercial value in marketing and operational scenarios. It leverages industry-leading multimodal data intelligence, enterprise knowledge graphs, and data privacy technologies to offer a suite of innovative application software. Key offerings include proprietary large models like DeepMiner for global commercial data intelligence analysis, AI-driven marketing solutions, operational intelligence tools, and comprehensive data management platforms. MININGLAMP also provides industry-specific solutions for automotive, retail, home furnishing, beauty, FMCG, and more, emphasizing data privacy and trusted intelligence.

The Open Japanese LLM Leaderboard is a platform designed for exploring and comparing large language models (LLMs) tailored for the Japanese language. Hosted on Hugging Face Spaces, this tool allows users to search for models by name, and apply filters based on type, size, and precision. It provides performance metrics and visualizations to help researchers, developers, and enthusiasts assess the capabilities of various Japanese LLMs. The leaderboard aims to facilitate the identification of top-performing models, supporting advancements in Japanese natural language processing and AI development. While the current live website indicates a runtime error, the intended functionality is to offer a comprehensive resource for evaluating and understanding the landscape of open Japanese LLMs.

506.ai offers a secure and flexible AI platform designed for businesses to seamlessly integrate AI assistants and solutions into their operations. The platform provides a dedicated AI infrastructure, enabling companies to automate processes, reduce costs, and ensure high data security. It supports intuitive access to AI through chat and seamless API integration into existing systems, usable on desktop or mobile. Key features include easy implementation, secure use of proprietary data with 100% GDPR compliance, and company-wide productivity enhancement through shared data storage and customizable AI assistants. 506.ai also offers consulting services, including AI introduction, implementation, and competency training, to ensure effective and sustainable AI adoption within organizations.

OpenCUA is a comprehensive open-source framework designed for scaling Computer-Use Agent (CUA) data and foundation models. It features AgentNet, the first large-scale computer-use task dataset spanning multiple operating systems and applications, and AgentNetTool, an annotation infrastructure for capturing human computer-use demonstrations. The framework also includes AgentNetBench, an offline evaluator for benchmarking model-predicted actions, and OpenCUA Models, end-to-end computer-use foundation models with strong planning and grounding capabilities. Notably, OpenCUA-72B achieves state-of-the-art performance on OSWorld-Verified, making it ideal for developers and researchers working on advanced AI agents.