Coding & Development

verl-agent is an extension of veRL specifically designed for training large language model (LLM) and vision-language model (VLM) agents using reinforcement learning (RL). It is the official code for the paper "Group-in-Group Policy Optimization for LLM Agent Training." A key differentiator is its step-independent multi-turn rollout mechanism, which allows for fully customizable per-step input structures, flexible history management, and modular memory. This design makes verl-agent highly scalable for very long-horizon, multi-turn RL training, such as tasks requiring up to 50 steps. The framework provides a diverse set of RL algorithms, including GiGPO, GRPO, and PPO, along with a rich suite of agent environments for both visual and text-based tasks, supporting models like Qwen3, LLaMA3.2, and LoRA fine-tuning.

Nova AI is an AI agent designed to transform QA by automating the generation and maintenance of end-to-end tests. It leverages generative AI to create tests based on real user behavior, ensuring comprehensive coverage that mirrors how users actually interact with an application. This approach helps eliminate redundancy, amplify efficiency, and reduce maintenance overhead by automatically updating test steps when UI elements or business logic change. Nova AI integrates with CI/CD pipelines, providing swift feedback directly in pull requests and supporting parallel execution across multiple environments. It offers actionable insights through dashboards, breaking down coverage by critical paths like Checkout or Sign Up, allowing teams to focus their efforts effectively. The platform supports browser-based and API testing, maintaining custom Chrome images and other major browsers for comprehensive coverage.

Outtake leverages agentic AI to secure modern attack surfaces, offering advanced search capabilities, real-time threat classification, and automated response. The platform is designed to protect the digital presence of brands by continuously scanning the open information environment for threats. It transforms noise into insights through Open Source Intelligence, monitoring emerging narratives, detecting threats from open sources, and protecting against location-based risks. For Digital Risk Protection, Outtake's AI agents automatically identify, prioritize, and dismantle impersonation attacks across various digital surfaces like social media, apps, and advertisements. Additionally, Outtake Verify is a browser extension that cryptographically signs emails and enables employees to report brand impersonations, helping to authenticate executive communications and prevent account takeovers.

Decisional is an AI agent platform designed to automate workflows by allowing users to build self-healing automation agents using natural language. Users describe their desired workflow in plain English, connect their applications, and the platform generates an agent that builds, tests, and runs the workflow 24/7. It simplifies complex operational tasks, offering a 10X simpler approach than traditional node-based workflow builders. Agents can read documents, use computers, edit spreadsheets, and call APIs, integrating seamlessly with tools like Gmail, Slack, and Stripe. The platform also provides visual workflow inspection, real-time output review in spreadsheets, and human feedback loops for necessary approvals.

ImagePrompt is a coding & development tool designed to facilitate AI image generation by converting images into text prompts. It also offers functionality to extract images from web pages, which can be useful for various purposes such as content creation, data collection, or design inspiration. The tool aims to streamline the process of generating effective prompts for AI models, making it easier for users to achieve desired visual outputs. Its dual capability of prompt generation from images and image extraction from web pages provides a versatile solution for those working with AI-driven visual content.

AiFlo is a comprehensive AI process automation platform designed to help businesses build enterprise applications and streamline operations at a significantly lower cost. It offers features like AI Process Automation, Low Code Workflow Management, and Integrations to connect various tools and automate complex processes. The platform includes an AI Co-Pilot for business users to automate tasks conversationally, and specialized AI tools for recruitment process automation, lead scoring, and email management. AiFlo aims to increase employee productivity by managing all business operations on a single platform, enabling digital transformation and competitive advantage through GenAI.

This repository offers PyTorch implementations of two significant transformer network architectures: "Attention is All You Need" by Vaswani et al. (NIPS 2017) and "Weighted Transformer Network for Machine Translation" by Ahmed et al. (Arxiv 2017). It serves as a valuable resource for researchers and developers interested in understanding, experimenting with, and applying transformer models in machine translation and other natural language processing tasks. The project is open-source, providing the full code for these implementations, making it accessible for academic study, practical development, and benchmarking against established models.

MindSpore is an open-source, self-developed AI framework by Huawei, designed to support full-scenario deep learning training and inference across device, edge, and cloud environments. It aims to provide a new AI programming paradigm, enabling developers to create efficient and flexible AI software and hardware applications. Key features include automatic differentiation based on source code transformation, automatic distributed parallel training, robust data processing capabilities, and a high-performance graph execution engine. MindSpore is primarily applied in AI fields such as computer vision and natural language processing, catering to data scientists and algorithm engineers. The framework fosters an AI development ecosystem through its open-source nature and community support, offering various toolkits for scientific computing, large models, and specific domains like CV and NLP.

Weave is an engineering intelligence platform designed to measure the value engineers derive from AI and to understand overall engineering work. It combines LLMs and domain-specific machine learning to normalize engineering output, benchmarked against expert standards rather than simple line counts. The platform separates human and AI contributions across pull requests, reviews, and deployments, offering insights into AI's impact on velocity, code quality, and review throughput. Weave connects various engineering tools—code, AI agents, project management, and communication—to unify data under one graph, attributing every contribution. It also provides features like AI health, AI ROI, AI usage, and agent observability, making it suitable for teams shipping with AI at scale.

Embedded LLM offers a no-code platform designed to help enterprises harness their data and build AI solutions rapidly. It simplifies the integration of Large Language Models (LLMs) into business workflows, allowing users to create powerful AI pipelines without coding. The platform features JamAI Base for collaborative AI and TokenVisor for managing and monetizing LLM workloads. Key capabilities include hardware acceleration optimized for AMD and multi-GPU clusters, robust data security with on-premise or private cloud deployment options, and full compatibility with the OpenAI API format. This ensures seamless migration and hybrid setups, preventing vendor lock-in and code rewrites. Embedded LLM aims to transform enterprise data into actionable AI solutions, boosting innovation and making advanced technology accessible to all business professionals.

unilm is an open-source project by Microsoft focused on large-scale self-supervised pre-training, encompassing a wide range of tasks, languages, and modalities. It provides foundational architectures like TorchScale, DeepNet, and Magneto, alongside a diverse collection of pre-trained models such as Kosmos, MetaLM, BEiT, WavLM, and LayoutLM. The project supports advancements in natural language processing, machine translation, speech recognition, document AI, and multimodal AI. It is designed for researchers and developers interested in building and experimenting with cutting-edge foundation models, offering resources for language, vision, and speech tasks, as well as multimodal applications.

Okareo offers a comprehensive platform for AI application developers to build and deploy high-quality LLM-powered applications. It enables users to simulate real users and automate agent behaviors with 'Drivers' to expose unexpected behaviors and edge cases. The platform allows for automated agent evaluation in CI/CD pipelines, catching failures early and ensuring reliable agents. Okareo also helps close the loop from production to development by pairing simulation with error tracking and synthetic data, generating new test cases and data for retraining or fine-tuning. This leads to faster fixes, stronger guardrails, and safer agents, providing full visibility and increasing accuracy for AI products.

Resolvd is an AI-powered platform designed to help healthcare supply chain teams identify and recover vendor pricing errors. It achieves this by scanning, extracting, and reconciling every invoice against every contract at the line-item level, creating a pricing source of truth. The platform deploys 'AI Workers' that act as digital staff, configured to existing systems and trained on specific processes to clear backlogs, validate data, and close loops that would otherwise require manual intervention. Resolvd aims to catch overpayments before payment, with health systems typically seeing measurable savings within the first quarter of deployment. It integrates with existing ERP environments like Workday, PeopleSoft, and Oracle, and document management systems such as OnBase and ImageNow, requiring minimal IT burden.

Datameister is an AI research and deployment lab based in Ghent, Belgium, focused on bridging advanced AI research with production engineering. They specialize in building and deploying systems for spatial and visual intelligence, encompassing computer vision, 3D generative AI, and MLOps. Their core capabilities include advanced computer vision for tracking, segmentation, and 2D-to-3D enrichment, 3D deep learning for processing meshes and point clouds, and 3D generative AI for creating context-aware 3D assets and synthetic training data. Datameister also develops scan-to-3D pipelines for virtual twins and physical AI solutions for robotic diagnostics. Their platform provides GPU-first infrastructure for visual AI workloads, integrating model development, deployment, monitoring, and EU compliance.

Skill Soup is an innovative platform designed for the emergent evolution of AI agent skills. It functions as an evolutionary arena where AI agents' capabilities compete, mutate, and evolve under community-driven selection pressure. Users can submit ideas for skills they want AI agents to build, and builders then compete to create them. The platform encourages active participation by allowing users to run builders, generate skills, and then use and vote on these skills to apply selection pressure. This 'survival of the fittest' environment ensures that better skills survive and evolve further, making it an experimental space for AI developers and researchers to test and improve AI agent functionalities.

Mezino Technologies specializes in leveraging AI and blockchain to transform business operations. They provide tailored solutions ranging from automating processes and enhancing security to building intelligent, future-ready platforms. Their approach involves a meticulous process of discovery, strategy, rapid prototyping, and iterative development, ensuring solutions are customized for specific market needs and growth plans. Mezino offers deep expertise in areas like predictive analytics, custom LLMs, conversational AI, smart contracts, tokenization, and secure ledgers. They focus on delivering scalable architecture, bulletproof security, and compliance, helping businesses transition from outdated systems to intelligent, decentralized solutions without complexity.

TheCodeWork is a comprehensive technology partner specializing in AI-powered software development and strategic consulting for startups and growing businesses. They offer end-to-end development of AI products, intelligent business platforms, and scalable mobile and web applications. Their services include AI strategy consulting, AI-driven process automation, product management, and integration of AI capabilities into existing enterprise systems. TheCodeWork focuses on building custom solutions tailored to specific business operations, rather than relying on templates, ensuring software fits the client's structure and long-term goals. They also provide strategic technology consulting to plan roadmaps, validate product ideas, and guide execution with full ownership.

Datalytica is a leading AI consulting and development firm dedicated to accelerating the adoption of artificial intelligence and emerging technologies across various sectors. They offer expertise in AI assurance and reverse engineering, as evidenced by their award from the AFRL Regional Network for groundbreaking work in these areas. Datalytica provides strategic and technical advisory support, helping organizations like ARPA-H shape research portfolios, refine program strategies, and strengthen execution to achieve measurable impact. Their work spans diverse applications, from operationalizing Large Language Models for the Department of Defense to digital modernization of industrial operations and advising on trusted AI for coalition-ready space superiority for NATO. Datalytica's insights are valuable for both government entities and the private sector, driving innovation and ensuring the effective deployment of AI solutions.

Huawei Cloud offers a robust and secure cloud platform designed to address industry challenges with AI-driven solutions. The service provides a wide array of cloud products, encompassing stable and reliable computing resources like elastic cloud servers and GPU-accelerated servers, secure storage solutions such as object storage and cloud disks, and advanced networking capabilities including virtual private clouds and elastic public IPs. Beyond infrastructure, Huawei Cloud integrates AI development platforms like ModelArts and AgentArts, enabling users to build and deploy intelligent applications. It also features MaaS (Model as a Service) for easy access to powerful large language models like GLM-5.1 and Kimi K2.6, facilitating AI innovation across various sectors. The platform emphasizes security with services like WAF and enterprise host security, ensuring data protection and compliance. Huawei Cloud aims to empower businesses with scalable, efficient, and intelligent cloud solutions for digital transformation.

Qoverse is an AI engineering company focused on developing and deploying production-ready AI agents tailored for specific business needs. Their expertise spans across finance, logistics, energy, and IT sectors, offering solutions from fraud detection to supply chain optimization. Qoverse provides specialized agents for tasks like Benford's Law analysis, Monte Carlo simulations, and LLM-powered compliance in finance, boasting 99.2% detection accuracy. They also build AI SaaS products like Founderr for solo founders, Homieye for PropTech tenant matching, and Rahnaward Academy for EdTech. Their process involves discovery, architecture, shipping, and scaling, ensuring robust and scalable AI solutions.

Promobot Corp. specializes in creating service robots designed to assist businesses across various sectors. Their robots feature autonomous navigation and advanced dialogue systems, powered by AI and neural networks. These intelligent machines can perform a range of tasks, including administrative assistance, medical support, consulting, tour guiding, concierge services, and promotional activities. Promobot robots are built for ease of use, allowing for simple customization of movements, emotions, and phrases without requiring specialized experience. They are designed for autonomous operation, reducing the need for constant human control, and come with a 1-year warranty and 7-day-a-week technical support. The company emphasizes safety with metal bodies and movement sensors to prevent harm, and offers various partnership options including discounts, leasing, and part payments.

CodeSleep is designed to significantly reduce AI coding costs, especially for users working with Claude Code. The tool achieves this by automating sequential development tasks, which helps optimize resource utilization and minimize manual intervention. It also facilitates the remote management of development tasks, offering flexibility and efficiency for distributed teams or individual developers working from various locations. A core aspect of CodeSleep's cost-saving strategy involves intelligently utilizing appropriate AI models for different tasks, ensuring that the most cost-effective yet capable AI is applied to each specific coding challenge. This strategic model selection helps in lowering overall AI expenditure while maintaining productivity.

Tabnine is an AI code assistant designed to accelerate and simplify software development for individuals and teams. It offers context-aware code suggestions, code completion, and AI agents for various stages of the SDLC, including planning, code creation, testing, and documentation. Tabnine prioritizes code privacy, security, and compliance, allowing deployment in cloud, on-premise, or air-gapped environments. It supports multiple LLMs, giving developers flexibility in their choice of models. The tool integrates with popular IDEs and can be fine-tuned on internal codebases, ensuring consistency and boosting productivity.

3LC.AI offers a comprehensive platform for AI data preparation and optimization, specifically designed for computer vision models. It illuminates the 'black box' of AI by combining labeling, debugging, and diagnosis capabilities. The tool seamlessly integrates into existing model training processes with minimal code changes, supporting popular frameworks like Python, Hugging Face, Ultralytics YOLO, Detectron2, PyTorch, and Jupyter. Clients have reported significant improvements, including a 30x reduction in false positives, a 50% increase in true positive rates, and a 75% reduction in training time, leading to reduced costs and CO2 emissions. 3LC.AI also allows users to keep their data in its current location, supporting Azure, Amazon, Google Cloud, network storage, and local storage.