An actionable framework to investigate and foster women inclusion in software development teams in proprietary software ecosystems

As consumer-grade GPUs have rapidly evolved, efforts have emerged to deploy these computational models for training and inference, typically handled by data center hardware. The paper explores optimization of two next-generation graphics computing units, the NVIDIA GeForce RTX 5090 and the AMD Radeon RX 9070, to optimize the new generation of ML and AI applications. We examine the internal compute pipelines, tensor/matrix acceleration capabilities, memory hierarchies, and software ecosystems (CUDA/cuDNN/TensorRT versus ROCm/MIOpen/HIP) that influence ML performance in a two-pronged architectural and empirical study. The convolutional networks, transformer models, diffusion architecture, and graph neural networks share a standard benchmarking model: training, inference latency, power consumption, precision scaling (FP32-INT8), and bottlenecks. The results of the experiment have demonstrated that the performance profiles of the RTX 5090 and the RX 9070 are different, i.e., the acceleration performance of mixed precision and kernel fusion is higher in the RTX 5090 as compared to the throughput performance of the RX 9070 in the BF16/INT8 workloads with the high memory-bandwidth utilization. Strategies for each platform. Platform-specific optimization strategies, such as kernel tuning, compiler optimization, memory prefetching, gradient checkpointing, and scaling to multiple GPUs, are developed and evaluated. Further, two case studies of real-world performance tuning of transformer fine-tuning and diffusion model inference are also presented. The findings highlight that hardware alone does not guarantee the best ML performance; effective optimization can deliver performance gains that are even more significant than raw compute alone. The paper will provide a step-by-step roadmap for practitioners, researchers, and engineers who may want to optimize the application of RTX 5090 and RX 9070 in artificial intelligence algorithms, as well as a future perspective on the standard models of unified programming on GPUs and emergent precision formats.

The LoongArch instruction set architecture (ISA) has become a cornerstone in efforts to build a secure, autonomous, and high-performance domestic computing ecosystem. To make Loongson processors practical for real software deployment, a dependable and well-optimized compiler is essential—particularly for emerging 32-bit platforms such as LoongArch32R. This study develops a complete and reproducible workflow for adapting and optimizing the GNU C Compiler (GCC) for LoongArch32R, enabling reliable instruction generation and performance-focused code transformation. The work combines several technical components: validation of the GCC backend, execution through QEMU in both user-level and system-level environments, incorporation of the MOS teaching operating system with custom benchmark applications, detailed examination of LSX SIMD auto-vectorization, and the introduction of a prototype custom vector instruction (VCUBE.W) through assembler-level extension. A structured benchmarking suite—including matrix multiplication, prime sieve, STREAM-like memory workloads, and memory operations—was implemented to evaluate optimization levels and compiler behavior. Performance measurements were analyzed and visualized using Python-based graphing tools. The experimental results show clear runtime improvements from standard optimization flags and demonstrate partial vectorization benefits, verifying that the ported compiler is functional, stable, and capable of generating efficient LoongArch32R code. Overall, the framework produced in this work offers a practical foundation for future compiler development, educational use, and broader software ecosystem support for LoongArch-based systems.

This article explores the role of unrecognised labour in corporate innovation systems via an analysis of researcher coding and discursive contributions to R, one of the largest statistical software ecosystems. Studies of online platforms typically focus on how platform affordances constrain participants’ actions, and profit from their labour. We innovate by connecting the labour performed inside digital platforms to the professional employment of participants. Our case study analyses 8,924 R package repositories on GitHub, examining commits and communications. Our quantitative findings show that researchers, alongside non-affiliated contributors, are the most frequent owners of R package repositories and their most active contributors. Researchers are more likely to hold official roles compared to the average, and to engage in collaborative problem-solving and support work during package development. This means there is, underneath the ‘recognised’ category of star researchers who transition between academia and industry and secure generous funding, an ‘unrecognised’ category of researchers who not only create and maintain key statistical infrastructure, but also provide support to industry employees, for no remuneration. Our qualitative findings show how this unrecognised labour affects practitioners. Finally, our analysis of the ideology and practice of free, libre and open source software (FLOSS) shows how this ideology and practice legitimate the use of ‘university rents’ by Big Tech. In conclusion, we argue that existing mechanisms are insufficient to ensure these digital commons’ sustainability: FLOSS needs broader systemic support.

Modern discovery is increasingly shaped by an integrated AI–Data–Compute–Governance stack that spans algorithms, software ecosystems, distributed infrastructure, cyber-physical systems, and socio-technical oversight. This review offers a pan-disciplinary synthesis across ten pillars Python/data-science ecosystems; ML/AI foundations including multimodality and RAG/agents; big data and the compute continuum (cloud–edge–HPC/quantum); IoT, robotics, and digital twins; bio/health informatics; geospatial/remote sensing; cybersecurity and privacy; blockchain for provenance; responsible governance; and sustainability/cost. We articulate a unifying lifecycle (design → train → evaluate → deploy → monitor → govern) and map cross-field patterns that consistently determine success: data quality over model size; retrieval-first, knowledge-integrated pipelines; agentic orchestration with strong evaluation; systems-level efficiency (compilers, quantization, distillation); privacy-by-design; and end-to-end assurance for safety, security, and robustness. Methodologically, we consolidate peer-reviewed literature and standards (2015–2025) from major digital libraries, structured via a transparent selection and evidence-grading protocol. The article contributes: (i) a taxonomy aligning methods and systems across scientific domains; (ii) reference blueprints for multimodal RAG/agent pipelines and edge-to-cloud deployment; (iii) comparative tables of tools, datasets, and evaluation suites; (iv) a measurement playbook spanning accuracy, reliability, security, privacy, latency, and energy/carbon; and (v) a 2025–2030 research roadmap prioritizing interpretable multi-agent systems, knowledge-grounded foundation models, privacy-preserving retrieval, green training/serving, and governance-aligned operations. By integrating perspectives from computer science, IT, data/AI engineering, and domain sciences, this review provides a coherent guide for researchers, practitioners, and policymakers seeki

High-quality data preprocessing is essential for untargeted metabolomics experiments, where increasing data set scale and complexity demand adaptable, robust, and reproducible software solutions. Modern preprocessing tools must evolve to integrate seamlessly with downstream analysis platforms, ensuring efficient and streamlined workflows. Since its introduction in 2005, the xcms R package has become one of the most widely used tools for LC-MS data preprocessing. Developed through an open-source, community-driven approach, xcms maintains long-term stability while continuously expanding its capabilities and accessibility. We present recent advancements that position xcms as a central component of a modular and interoperable software ecosystem for metabolomics data analysis. Key improvements include enhanced scalability, enabling the processing of large-scale experiments with thousands of samples on standard computing hardware. These developments empower users to build comprehensive, customizable, and reproducible workflows tailored to diverse experimental designs and analytical needs. An expanding collection of tutorials, documentation, and teaching materials further supports both new and experienced users in leveraging broader R and Bioconductor ecosystems. These resources facilitate the integration of statistical modeling, visualization tools, and domain-specific packages, extending the reach and impact of xcms workflows. Together, these enhancements solidify xcms as a cornerstone of modern metabolomics research.

Computational modeling of chemical reactions is fundamental to modern synthetic chemistry but is often hindered by a fragmented software ecosystem and the complexity of accurately representing the reaction mechanisms. To address this, we introduce SynKit, an open-source Python library that provides a unified, chemically intuitive framework for reaction informatics. SynKit performs core tasks such as reaction canonicalization and transformation classification, while other functionalities─such as synthetic route construction through rule composition─are supported through integration with external libraries. The newly introduced Mechanistic Transition Graph extends the traditional net-change representation of the Imaginary Transition State by explicitly modeling the sequence of bond-forming and bond-breaking events, capturing transient intermediates, and providing deeper mechanistic insight. Designed for easy installation and broad compatibility, SynKit integrates smoothly into existing computational workflows for exploring complex Chemical Reaction Networks. For more advanced network analyses, it interfaces with specialized tools (e.g., MØD) to support exhaustive mechanism enumeration and kinetics-aware studies. By combining advanced mechanistic modeling with an accessible, modular design, SynKit supports more reproducible and rigorous research in automated synthesis planning.

Recent progress in distributed AI systems intensifies the demand for reliable coordination, unified context exchange, and efficient communication among autonomous analytical agents operating across dynamic, heterogeneous environments. This study analyzes these challenges and introduces the Model Context Protocol (MCP) as a structured and extensible solution that enhances context-driven interaction within distributed software ecosystems. The proposed framework integrates layered context exchange, persistent memory structures, deterministic communication channels, and standardized interaction patterns, aligning independent analytical models into a cohesive system capable of real-time information acquisition, multi-stage task execution, and seamless integration of structured knowledge from diverse and continuously evolving data sources. MCP strengthens cooperative and system-level reasoning by enabling agents to synchronize memory states, maintain contextual continuity, and dynamically adjust analytical workflows as operational conditions evolve, even under high variability and unpredictable workloads. The architecture incorporates automated task decomposition, adaptive memory alignment, context persistence, robust tool invocation, and flexible routing logic, supporting hybrid deployments across edge devices, local clusters, and cloud infrastructures while maintaining high reliability and analytical stability. The study evaluates MCP through federated learning simulations, distributed workflow experiments, and comparative benchmarks, demonstrating consistent improvements in inference latency, throughput efficiency, policy adaptability, context-aware decision-making, and multi-agent collaboration across complex distributed environments. These results confirm MCP's ability to reduce integration overhead, mitigate data fragmentation, stabilize communication pathways, enhance analytical robustness, and improve accuracy in both small-scale and enterprise-level scenarios. Consequently, MCP establishes a scalable foundation for resilient analytical pipelines and significantly expands the practical adoption of autonomous agents in enterprise automation, distributed decision-support systems, and next-generation cognitive infrastructures.

The semiconductor industry is the core of the digital information age and the technological linchpin of global connectivity. Its technological innovation and application expansion represent the cutting edge of contemporary scientific and technological advancement. In recent years, with the rise and proliferation of artificial intelligence technologies, particularly generative AI, the global semiconductor technology industry is undergoing unprecedented transformation, giving rise to numerous industry giants and emerging players. Among these industry leaders, NVIDIA has leveraged its years of technological accumulation and strategic positioning in the GPU field to gain significant competitive advantages. It has emerged as a core participant in the new wave of AI, surging to become the world’s most valuable company by market capitalization. Its development trajectory and market performance have drawn widespread societal attention. This article focuses on NVIDIA’s data center business, analyzing its development trajectory and the characteristics of the AI chip industry. It employs a SWOT model to examine NVIDIA’s internal strengths and external environment, concluding with a summary of NVIDIA’s performance and financial status, providing an objective overview of its market position and current development. Through this analysis, NVIDIA is expected to maintain its market position by sustaining R&D investment, deepening its software ecosystem, optimizing supply chain management, and proactively addressing geopolitical challenges. This strategic approach positions the company to sustain high-growth momentum in the increasingly competitive AI chip market.

This study investigates the impact of AI- and ML-driven Predictive Quality Orchestration (PQO) on enhancing test intelligence, defect forecasting, and compliance optimization within Agile DevOps environments applied to U.S. healthcare and Human Resource Management (HRM) systems. Employing a quantitative, predictive analytical approach, data were collected from five major organizations integrating AI-enabled DevOps practices. Using machine learning algorithms; Random Forest (RF), Long Short-Term Memory (LSTM), and Gradient Boosting Machine (GBM) the study developed and validated predictive models for quality orchestration. Results revealed that GBM achieved the highest predictive performance (accuracy = 94.5%, ROC-AUC = 0.96), while healthcare systems demonstrated superior test coverage, lower defect density, and faster resolution rates compared to HRM systems. Regression analysis confirmed significant positive relationships between AI Model Complexity, Data Quality Index, and Agile Process Maturity with key performance outcomes. Post-implementation, compliance deviation reduced by 61%, and audit readiness improved by 25.9%. These findings underscore that PQO not only improves software reliability and compliance assurance but also establishes a self-learning framework that continuously optimizes performance in critical, regulated environments. The study concludes that integrating AI-driven orchestration into DevOps pipelines is a strategic pathway to achieving sustainable, intelligent, and compliant software ecosystems.

Exploring developer experience factors in software ecosystems

Overcoming experimentation challenges in software ecosystems of large product and service organizations: A participatory action research study

Code signing is a fundamental tool for establishing trust and integrity in today's software ecosystems. However, as traditional code-signing methods are largely static, these methods only provide a guarantee based on the certificate issuer and validity at the time of signing, without regard to the contextual behavior associated with the signing process. As such, they are vulnerable to various factors, such as insider abuse, unauthorized access to signing keys, and post-signing tampering. This paper presents a novel paradigm - Leveraging Generative AI for Intelligent Code Signing and Tamper Detection - which is built on an AI-based Trust Graph framework. The proposed work utilizes Generative Artificial Intelligence (Generative AI) and Graph Neural Networks (GNNs) to model the dynamic relationships between the developers, the repositories, the certificates, and the binaries, allowing for the detection of abnormal signing behavior patterns that may indicate compromise. The system continuously analyzes behavioral patterns as well as provenance-type data to achieve AI-based trust scoring and contextual anomaly detection to uncover instances of unauthorized use of keys, insider tampering, and subtle compiler-related or code modification instances, which are typically missed by static methods. The adaptive trust ecosystem enhances both the integrity of the software and resilience in the software supply chain, and clearly demonstrates the way in which generative AI can bridge the gap between traditional authenticity verification and real-time threat mitigation.

SUMMARYMillions of SARS-CoV-2 genome sequences were collected during the COVID-19 pandemic, forming a dataset of unprecedented richness. Estimated genealogies are fundamental to understanding this ocean of data and form the primary input to many downstream analyses. A basic assumption of methods to infer genealogies from viral genetic data is that recombination is negligible and the genealogy is a tree. However, recombinant lineages have risen to global prevalence, and simple tree representations are therefore incomplete and potentially misleading. We present sc2ts, a method to infer reticulate genealogies as an Ancestral Recombination Graph (ARG) in real time at pandemic scale. We infer an ARG for 2.48 million SARS-CoV-2 genomes, which leverages the widely used tskit software ecosystem to support further analyses and visualisation. This rich and validated resource clarifies the relationships among recombinant lineages, quantifies the rate of recombination over time, and provides a lower bound on detectable recombination.

ABSTRACT - Contemporary academic environments struggle with operational inefficiencies stemming from disparate information repositories, lack of actionable intelligence on student trajectories, and siloed communication channels across institutional stakeholders. This paper introduces Campus Connect, a comprehensive software ecosystem that consolidates administrative operations and applies predictive analytics to forecast student academic outcomes. Leveraging the MERN technology stack combined with a Flask-based machine learning microservice, the platform delivers differential user interfaces for students, faculty, and administrators with role-specific visualizations and controls. Performance prediction employs an optimized Random Forest ensemble method analyzing attendance patterns, continuous assessment scores, cumulative performance indices, and course credit completion. Technical validation demonstrates model effectiveness at 89.33% accuracy, sub-2-second latency across user interactions, and horizontal scalability supporting simultaneous sessions exceeding 2000 concurrent participants. This contribution advances educational data mining through an integrated, production-ready implementation bridging institutional data governance, contemporary full-stack architecture, and automated predictive intelligence for timely intervention with struggling learners. · Keywords: Educational Management Platform, Predictive Analytics, Student Academic Forecasting, MERN Technology Stack, Ensemble Learning Methods, Data-Driven Academic Support, Multi-Role Web Application

This article reports on the results of the 13th ACM/IEEE International Workshop on Software Engineering for Systems-of-Systems and Software Ecosystems (SESoS 2025) in which researchers and practitioners discussed ideas and experiences on the research and practice for the development and evolution of complex softwareintensive systems, more specifically systems-of-systems (SoS) and software ecosystems (SECO). SESoS 2025 was co-located with the 47th IEEE/ACM International Conference on Software Engineering (ICSE 2025). After over a decade running this workshop, the SESoS community is advancing on how to cope with the different dimensions that should be considered in the engineering of those classes of systems (i.e. technological, organizational, and social), and also is taking awareness of newer challenges for inclusiveness and sustainability. In addition, benchmarks for conducting research on the areas as well as approaches for investigating emerging domains (smart ecosystems) and non-functional requirements on those systems were pointed out as relevant challenges.

The China Seismo Electromagnetic Satellite (CSES) mission provides in situ measurements of the electromagnetic field, plasma, and charged particles in the topside ionosphere. Each CSES spacecraft carries several different scientific payloads delivering a wealth of information about the ionospheric plasma dynamics and properties, as well as measurement about energetic particles precipitating in the ionosphere. In this work, we introduce CSESpy, a Python package designed to provide an interface to CSES data products, with the aim of easing the pathway for scientists to carry out analyses of CSES data. Beyond simply being an interface to the data, CSESpy aims to provide higher-level analysis and visualization tools, as well as methods for combining concurrent measurements from different instruments, so as to allow multipayload studies in a unified framework. Moreover, CSESpy is designed to be highly flexible as such, it can be extended to interface with datasets from other sources and can be embedded in wider software ecosystems. We highlight some applications, also demonstrating that CSESpy is a powerful visualization tool for investigating complex events involving variations across multiple physical observables.

Abstract The pull-based development model facilitates global collaboration within open-source software projects. However, whereas it is increasingly common for software to depend on other projects in their ecosystem, most research on the pull request decision-making process explored factors within projects, not the broader software ecosystem they comprise. We uncover ecosystem-wide factors that influence pull request acceptance decisions. We collected a dataset of approximately 1.8 million pull requests and 2.1 million issues from 20,052 GitHub projects within the NPM ecosystem. Of these, $$98\%$$ depend on another project in the dataset, enabling the study of collaboration across dependent projects. We employed social network analysis to create a collaboration network in the ecosystem, and mixed-effects logistic regression and random forest techniques to measure the impact and predictive strength of the tested features. We find that gaining experience within the software ecosystem through active participation in issue-tracking systems, submitting pull requests, and collaborating with pull request integrators and the ecosystem community benefits all open-source contributors, especially project newcomers. These results are complemented with an exploratory qualitative analysis of 538 pull requests. We find that developers with ecosystem experience make contributions more commonly associated with mature developers. For example, they introduce new features and bug fixes less commonly than dependency updates as part of maintenance. Zooming in on a subset of 111 pull requests with clear ecosystem involvement, we find 3 overarching and 10 specific reasons why developers involve ecosystem projects in their pull requests. For example, when another project has implemented a solution that can be used as a reference implementation. The results show that combining ecosystem-wide factors with features studied in previous work to predict the outcome of pull requests reached an overall F1 score of 0.92. However, the outcomes of pull requests submitted by newcomers are harder to predict. Our study identified some benefits associated with ecosystem-wide collaboration dynamics, laying the groundwork for future work in this direction.

Advancements in methodologies for efficient large-scale acquisition of high-resolution serial microscopy image data have opened new possibilities for experimental studies of cellular and subcellular features across whole brains in animal models. There is a high demand for open-source software and workflows for automated or semi-automated analysis of such data, facilitating anatomical, functional, and molecular mapping in healthy and diseased brains. These studies share a common need to consistently identify, visualize, and quantify the location of observations within anatomically defined regions, ensuring reproducible interpretation of anatomical locations, and thereby allowing meaningful comparisons of results across multiple independent studies. Addressing this need, we have developed a suite of desktop and web-applications for registration of serial brain section images to three-dimensional brain reference atlases (QuickNII, VisuAlign, WebAlign, WebWarp, and DeepSlice) and for performing data analysis in a spatial context provided by an atlas (Nutil, QCAlign, SeriesZoom, LocaliZoom, and MeshView). The software can be utilized in various combinations, creating customized analytical pipelines suited to specific research needs. The web-applications are integrated in the EBRAINS research infrastructure and coupled to the EBRAINS data platform, establishing the foundation for an online analytical workbench. We here present our software ecosystem, exemplify its use by the research community, and discuss possible directions for future developments.

Microbial co-occurrence network inference is often hindered by low accuracy and tool dependency. We introduce microbetag, a comprehensive software ecosystem designed to annotate microbial networks. Nodes, representing taxa, are enriched with phenotypic traits, while edges are enhanced with metabolic complementarities, highlighting potential cross-feeding relationships. microbetag’s online version relies on microbetagDB, a database of 34,608 annotated representative genomes. microbetag can be applied to custom (metagenome-assembled) genomes via its stand-alone version. MGG, a Cytoscape app designed to support microbetag, offers a streamlined, user-friendly interface for network retrieval and visualization. microbetag effectively identified known metabolic interactions and serves as a robust hypothesis-generating tool.Supplementary InformationThe online version contains supplementary material available at 10.1186/s13059-025-03769-2.