As general-purpose LLMs become widely accessible, tailoring their outputs to domain-specific tasks remains challenging. This paper investigates how custom, pattern-guided datasets and fine-tuning can specialize LLM behavior. We run a four-round open-vs-closed comparison: (1) evaluate commercially hosted models with an open-source dataset; (2) repeat with a synthetic, pattern-aligned variant; (3) replicate both rounds with open-source LLMs; and (4) demonstrate a real-world case by fine-tuning an open model to answer WordPress technical support queries using a finalized, pattern-constrained dataset. Across rounds, fine-tuned open-source models achieved performance comparable to closed-source baselines while offering greater controllability and deployment flexibility. We detail dataset design, the role of response patterns in producing consistent outputs, and practical trade-offs observed during fine-tuning and real-world use. The findings indicate that with well-crafted custom datasets, open-source LLMs can be reliably specialized for targeted applications without reliance on proprietary APIs.

In software engineering, the integration of machine learning (ML) and artificial intelligence (AI) components into modern web services has become commonplace. To comply with evolving regulations, such as the EU AI Act, the development of AI models must adhere to the principles of transparency. This includes the training data used, the intended use, potential biases, and the risks associated with these models. To support these goals, documents named Model Cards were introduced to standardize ethical reporting and allow stakeholders to evaluate models based on various goals. In our ongoing research, we aim to automate risk analysis and regulatory compliance checks in software systems. We envision that model cards can serve as useful tools to achieve the goal. Given the evolving format of model cards over time, we conducted a state-of-the-art review of the current state and practice of model cards by analyzing 90 model cards from four model repositories to assess their relevance to our vision. The study's contribution is a thorough analysis of the model cards' structure and content, as well as their ethical reporting. Our study reveals the variance in information reporting, the loose structure, and the lack of ethical reporting in the model cards. Based on the findings, we propose a unified model card template that aims to enhance the structure, promote greater transparency, and establish a foundation for future machine-interpretable AI model cards.

There is an increasing need for practitioners to address legislative and ethical issues in both the development and deployment of data-driven applications with AI/ML due to growing concerns and regulations, such as GDPR and the EU AI Act. Thus, the field needs a systematic framework for assessing risks and helping to stay compliant with regulations in designing and deploying software systems. Clear and concise descriptions of risks associated with each model and data source are needed to guide the design without acquiring deep knowledge of the regulations. In this paper, we propose a reference architecture for an ethical orchestration system that manages distributed AI/ML applications on the cloud–edge continuum and present a proof-of-concept implementation of the main ideas of the architecture. Our starting point is the methods already in use in the industry, such as model cards, and we extend the idea of model cards to data source cards and software component cards, which provide practitioners and the automated system with relevant information in actionable form. With the metadata card based orchestration system and information about the risk levels of the target infrastructure, the users can create deployments of distributed AI/ML systems that fulfill the regulatory and other requirements.

In the era of 6G, developing and managing software requires cutting-edge software engineering (SE) theories and practices tailored for such complexity across a vast number of connected edge devices. Our project aims to lead the development of sustainable methods and energy-efficient orchestration models specifically for edge environments, enhancing architectural support driven by AI for contemporary edge-to-cloud continuum computing. This initiative seeks to position Finland at the forefront of the 6G landscape, focusing on sophisticated edge orchestration and robust software architectures to optimize the performance and scalability of edge networks. Collaborating with leading Finnish universities and companies, the project emphasizes deep industry-academia collaboration and international expertise to address critical challenges in edge orchestration and software architecture, aiming to drive significant advancements in software productivity and market impact.

Agile software companies applying the DevOps approach require collaboration and information sharing between practitioners in various roles to produce value. Adopting new development practices affects how practitioners collaborate, requiring companies to form a closer connection between business strategy and software development. However, the types of information management, sales, and development needed to plan, evaluate features, and reconcile their expectations with each other need to be clarified. To support practitioners in collaborating and realizing changes to their practices, we investigated what information is needed and how it should be represented to support different stakeholders in their tasks. Compared to earlier research, we adopted a holistic approach—by including practitioners throughout the development process—to better understand the information needs from a broader viewpoint. We conducted six workshops and 12 semi-structured interviews at three Finnish small and medium-sized enterprises from different software domains. Thematic analysis was used to identify information-related issues and information and visualization needs for daily tasks. Three themes were constructed as the result of our analysis. Visual information representation catalyzes stakeholder discussion, and supporting information exchange between stakeholder groups is vital for efficient collaboration in software product development. Additionally, user-centric data collection practices are needed to understand how software products are used and to support practitioners’ daily information needs. We also found that a passive way of representing information, such as a dashboard that would disturb practitioners only when attention is needed, was preferred for daily information needs. The software engineering community should consider reviewing the information needs of practitioners from a more holistic view to better understand how tooling support can benefit information exchange between stakeholder groups when making product development decisions and how those tools should be built to accommodate different stakeholder views.

PhD programmes run by universities are planned for the needs of full-time students who have funding for the study period and an opportunity to work full time as a member of the academic community. This approach aims at following the optimal way to the goals set by the State and the university for these studies - young students, immediate continuation after MSc graduation, with membership in the university's PhD programme ensuring progress. We call these students “administrative optimal” PhDs. However, universities also have a number of postgraduate students that work in industry. These students work full time, have significant and useful experience from practice, but have more limited resources to use for PhD-related activities. This paper reports on the challenges related to PhD students working in industry based on experiences from the field of ICT; the root cause of which is the missing support provided by the university, which focuses on “administrative optimal” students in their PhD programmes. The paper focuses on ICT field, because it is our experience base. We believe that in ICT area it is more common to work simultaneously to studies. This fits especially in MSc studies in Finland, but the same tradition continues even in PhD level. The studies of industrial students typically need twice the time and encounter a variety of problems having their roots in the mental, practical and motivational side. We list the challenges and handle them in a structured way in the form of anti-patterns. To provide a context for our findings we have included a discussion related to general aspects of PhD studies in the paper. The aim of the paper is to open up the discussion about this important topic and report our summarised “close to one hundred years” of experience in mentoring and supervising PhD students.

Smart cities are getting more and more attention due to urbanization and IoT trends. At the same time, 360-degree videos are also getting more popular. The watchers of 360-degree videos provide a data source that fit to the data collection aim of smart cities. This paper explores how well 360-degree video user data canbe collected, using MQTT as a data transfer protocol, and analyzed with an open source IoT platform. The results suggest that using MQTT with the chosen IoT platform is convenient and general chart visualizations can provide useful insight about 360-degree video watchers. The used research method is design science.

Decomposition is one of the most complex tasks during the migration from monolithic systems to microservices, generally performed manually, based on the experience of the software architects. In this work, we propose a 6-step framework to reduce the subjectivity of the decomposition process. The framework provides software architects with a set of decomposition options, together with a set of measures to evaluate and compare their quality. The decomposition options are identified based on the independent execution traces of the system by means of the application of a process-mining tool to the log traces collected at runtime. We validated the process, in an industrial project, by comparing the proposed decomposition options with the one proposed by the software architect that manually analyzed the system. The application of our framework allowed the company to identify issues in their software that the architect did not spot manually, and to discover more suitable decomposition options that the architect did not consider. The framework could be very useful also in other companies to improve the quality of the decomposition of any monolithic system, identifying different decomposition strategies and reducing the subjectivity of the decomposition process. Moreover, researchers could extend our approach increasing the support and further automating the decomposition support.

Background. The need to release our products under tough time constraints has required us to take shortcuts during the implementation of our products and to postpone the correct implementation, thereby accumulating Technical Debt. Objective. In this work, we report the experience of a Finnish SME in managing Technical Debt (TD), investigating the most common types of TD they faced in the past, their causes, and their effects. Method. We set up a focus group in the case-company, involving different roles. Results. The results showed that the most significant TD in the company stems from disagreements with the supplier and lack of test automation. Specification and test TD are the most significant types of TD. Budget and time constraints were identified as the most important root causes of TD. Conclusion. TD occurs when time or budget is limited or the amount of work are not understood properly. However, not all postponed activities generated ”debt”. Sometimes the accumulation of TD helped meet deadlines without a major impact, while in other cases the cost for repaying the TD was much higher than the benefits. From this study, we learned that learning, careful estimations, and continuous improvement could be good strategies to mitigate TD These strategies include iterative validation with customers, efficient communication with stakeholders, meta-cognition in estimations, and value orientation in budgeting and scheduling.

Understanding users, their requirements and usage patterns helps building better software. To continuously improve operations, user-interaction (U-I) data can provide developers with interesting new possibilities. This study aims at gaining an understanding of how software teams can start the use of U-I data and what challenges they face with it. In this paper, we describe three cases from different organizations. This includes explaining the activities, objectives, and challenges of each team in their efforts to begin using U-I data. We conducted the study with Action Design Research (ADR) method, resulting in findings from each case by intervening in the work of these teams. As a contribution, we designed a U-I data utilization method that summarizes teams' activities. Secondly, we refined and validated the categorizations of U-I data analysis and utilization objectives, and finally we categorized the challenges that the case teams faced. Together, these contributions lay out clear steps also for other software teams in starting the utilization of U-I data.

As Internet‐of‐Things (IoT) devices become more powerful, they can also become full participants of Internet architectures. For example, they can consume and provide RESTful services. However, the typical network infrastructures do not support the architecture and middleware solutions used in the cloud‐based Internet. The authors show how systems designed with RESTful architecture can be implemented by using an IoT‐specific technology called message queuing telemetry transport (MQTT). Their example case is an application development and deployment system that can be used for remote management of IoT devices. To evaluate the proposed solution, they performed resource consumption experiments to compare HTTP and request‐response usage of MQTT. The results suggest that MQTT uses less central processing unit time and memory.

Depending on one's viewpoint, a generic standards-compatible web browser supports three, four or five built-in application rendering and programming models. In this paper, we provide an overview of the built-in client-side web application architectures. While the dominance of the base HTML/CSS/JS technologies cannot be ignored, we foresee Web Components and WebGL gaining popularity as the world moves towards more complex and even richer web applications, including systems supporting virtual and augmented reality.

Modern software development is supported by a rich set of tools that accumulate data from the software process automatically. That data can be used for understanding and improving software processes without any manual data collection. In this paper we introduce an industrial case where data visualization of issue management system was used to investigate software projects. The results of the study show that visualization of issue management system data can really reveal deviations between planned process and executed process.

Management of technical debt is a challenging and poorly understood task, and it is becoming even harder in the case of modern software engineering practices like Agile development and Continuous Delivery. In this research we assume an agile software development and management process where the organization selects the tasks in the beginning of each sprint. The candidate tasks include implementation of new features with assumed business value and paying back technical debt. The organization needs to select a combination of tasks that is implementable by the available resources and maximize the benefit for the organization. The required optimization problem in a large project is complex and is also a multi-objective problem, which involves trade-off between short-term and long-term value delivered by the software. In this paper, we apply a multiobjective genetic algorithm for solving such an optimization problem. The potential of the algorithm is demonstrated by applying it to a student project.

In the field of improving software processes, one of the clear trends has been the ever tighter engagement of end users in the software development process. This is demonstrated by the shift from Agile processes to Continuous Deployment, which requires more rapid ways to validate the developed software and its value than is possible with traditional communication mechanisms and methods, such as face to face conversations with customers. While post-deployment data has been used for years as an extra data source - companies like Microsoft and Intuit have moved a few steps further from that already - we believe that there are numerous uncovered ways of taking advantage of post-deployment data in software development. In this paper, we study how automatically collected post-deployment data could be used for responding to knowledge needs of software development teams. The paper builds on data collected from a number of companies operating in Finland using a questionnaire study. The focus of questionnaire study was to approach post-deployment data - especially usage data - as means of getting information to support understanding of customer and end users.

Software development methods are shifting towards faster deployments and closer to the end users.Their ever tighter engagement of end-users also requires new technologies for gathering feedback from those users.At the same time, widespread Internet connectivity of different application environments is enabling the collection of this post-deployment data also from sources other than traditional web and mobile software.However, the sheer number of different alternatives of collecting technologies makes the selection a complicated process in itself.In this paper, we describe the process of data-driven software development and study the challenges organizations face when they want to start guiding their development towards it.From these challenges, we extract evaluation criteria for technological approaches to usage data collecting.We list such approaches and evaluate them using the extracted criteria.Using a design science approach, we refine the evaluation criteria to a selection framework that can help practitioners in finding a suitable technological approach for automated collecting of usage data.

The liquid metaphor refers to software that operates seamlessly across multiple devices owned by one or multiple users. Liquid software architectures can dynamically deploy and redeploy stateful software components and transparently adapt them to the capabilities of heterogeneous target devices. The key design goal in liquid software development is to minimize the efforts that are related to multiple device ownership (e.g., installation, synchronization and general maintenance of personal computers, smartphones, tablets, home displays, cars and wear-able devices), while keeping the users in full control of their devices, applications and data. In this paper we present a design space for liquid software, categorizing and discussing the most important architectural issues and alternatives. These alternatives represent relevant capabilities offered by emerging technologies and deployment platforms that are then positioned and compared within the design space presented in the paper.

Global software engineering (GSE) has become common in the software industry. Distributed development work comes with many challenges, especially related to communication and coordination. Thus, it is essential to also teach and prepare the new population of software engineers to be aware of - and familiar with - these hurdles. We addressed this need by arranging a joint (agile) software project course between universities in Finland and Norway. We had three teams - one with students from both countries and two with students from the Norwegian university only. The students were given a teacher mentor, a handbook on agile practices and state-of-the-art tools for project management and software development. Apart from monitoring the teams' progress, we also collected data from emails, questionnaires, student reports and interviews during and after the project. The main lesson learned is that the global and local teams are mostly facing the same challenges - especially when it comes to team building, clear project roles, and communication and management issues. Ultimately, our results show that the challenges were harder to solve by the global team and that not solving challenges in a timely manner had more serious consequences.

Software process data is available in several tools such as version control systems, issue trackers and test and build systems to name a few. Using the data gathered in these software engineering tools would be ideal for collecting different kinds of software processes and product metrics as the data is already automatically gathered by the tools. However, the tools present and store the data in various formats. The data collection methods and interfaces also vary between the tools. This closes the software engineering data into silos and makes it hard to build reusable analysis and visualizations for the data. In this position paper we present a unified model for software engineering data and a framework for data collection, conversion and storing that utilizes our model. The aim of the model is to define a common format for software engineering data which is not dependent on specific software engineering tools or the software engineering process and thus can be used as a basis for building reusable visualization and analysis components. To demonstrate that we can build reusable visualization plugins on top of the framework, we created a timeline visualization plugin. The visualization plugin is used to visualize two data sets from industrial software projects that have different contexts and semantics.

Today, the digital life of people in developed markets is dominated by PCs and smart phones. Yet, as successful as PCs and smart phones are, the dominant era of PCs and smart phones is about to come to an end. Device shipment trends indicate that the number of web-enabled devices other than PCs and smart phones will grow rapidly. In the near future, people will commonly use various types of internet-connected devices in their daily lives. Unlike today, no single device will dominate the user's digital life. The transition to a world of multiple device ownership is still rife with problems. Since devices are mostly standalone and only stay in sync in limited ways, the users will have to spend a lot of time managing them. These device management chores become much more tedious as the number of devices in a person's life increases. In this paper we look at the ongoing paradigm shift towards multiple device ownership and its implications for software architecture. We argue that the transition to multiple device ownership will eventually lead us to liquid software - an approach that will allow data and applications to seamlessly move between multiple devices and screens. The new era will imply significant changes in the development, deployment and use of software, opening up new opportunities in software engineering research as well.

In the current web, end-users' involvement on creating content for the modern web applications and participation on tailoring their own web applications is increasing. These applications are often called Personalized Web Tasks. Thus, in addition to software developers, end-users are now participating in the service development through personal service configurations. End-user driven development poses new requirements on the development tools, which traditionally are targeted for software developers. We discuss the requirements for end-user driven integration of data-oriented web resources and use our own prototype tool as an example implementation. We describe generic data types for web services and propose a technique for adding metadata to service descriptions. The metadata is used to help the user to build service configurations and to convert data between different formats used by web services.