Innovation and Software

Optimization of Large Language Models (LLMs) through Prompt Engineering

Crishtian Brenon Paz Fernández — 2025-09-30

This article explored the impact of prompt engineering on optimizing the performance of large language models (LLMs) such as GPT and BERT. Prompt engineering was introduced as an innovative approach that involved designing specific instructions to guide the models' responses, enhancing their accuracy and relevance without modifying their internal parameters. The study evaluated methodologies for constructing effective prompts, compared different strategies such as few-shot and zero-shot learning, and analyzed practical cases in areas like text generation, question answering, and sentiment analysis. The results demonstrated that a strategic design of prompts could significantly improve response quality, reduce errors, and expand the range of LLM applications.

Web application for classifying and assisting in incident management using OpenAI LLMs

Diego Sebastian Vasquez Jaramillo — 2025-09-30

The proposal for a web application to assist in the management of technical incidents in the field of information technology is established. The implementation was carried out with a 3-layer architecture, based on web technologies using React, Laravel, and a relational database. Large language models were implemented, applying instruction design techniques to analyze descriptions of technical incidents and automatically provide suggestions and classify priority, based on criteria for incidents generated in the present. The proposal was developed based on the SCRUM agile methodology and validated with real users, who evaluated the functionality and accuracy of the system. The tool achieved a 77.3% accuracy in proposing correct suggestions, excelling in categories such as software and networks. These results demonstrated the usefulness of the solution as support in the selection of solutions and in reducing cognitive effort during the initial stages of diagnosis. It is concluded that the use of LLMs in technical support represents an effective alternative for optimizing processes, as long as it is used as a complement to human experience.

Impact of Information and Communication Technologies (ICT) on Supply Chain Integration in Construction Projects

Sergio Quispe — 2025-09-30

In recent years, the construction sector has undergone a significant transformation driven by digitalization. In this context, the use of Information and Communication Technologies (ICT) as a means of optimizing supply chain management (SCM) has gained particular relevance. These technologies have not only proven useful in improving traceability and efficiency in construction processes, but have also contributed to strengthening communication between the various stakeholders involved throughout the project lifecycle. Several recent studies have highlighted the benefits of digital tools such as BIM (Building Information Modeling), the Internet of Things (IoT), blockchain, and platforms.

Collaborative technologies, which enable more integrated and transparent project management (Wu et al., 2022; Celik et al., 2023). This systematic review, based on the analysis of 30 scientific articles indexed between 2019 and 2024, shows that the application of these technologies favors interorganizational coordination, reduces the likelihood of errors, and increases the system's capacity to adapt to changes and unforeseen events (Hargaden et al., 2021; Fernandez-Carames et al., 2024). However, several challenges are also identified that hinder their implementation. These include a lack of common standards, resistance to change on the part of teams, high initial costs, and limited training in digital skills. Despite these barriers, researchers agree that establishing appropriate regulatory frameworks and strengthening digital governance can facilitate a more widespread adoption of ICTs in construction projects. This would allow to fully exploit its potential in the improvement and integration of the supply chain (Qian & Papadonikolaki, 2020; Sharma et al., 2023).

Design of a microservices architecture in virtual containers

Samantha Yazmin Elizalde Valencia — 2025-09-30

This article presents the methodological proposal for the design and implementation of a containerized microservices architecture. The proposal covers the methodologies used for the research phase of the initial stages of project development, as well as the methodology used to create a test computing system. An architecture based on domain architectures is proposed, as well as the design of the system's physical architecture, based on the requirements presented by the collaborating institution for the test case. The results describe the configuration of the microservices and containers, as well as their integration into a common network of running services.

Intelligent automation for successful processes: n8n as a driver of digital efficiency

Mario S. Moreno — 2025-09-30

In a context where digital systems are becoming increasingly fragmented, process integration and automation are no longer a technical luxury, but a structural necessity for scaling, protecting data, and improving the end-user experience. This paper analyzes the strategic potential of n8n, an open-source automation and integration platform that combines the best of a low-code approach with the extensibility of traditional development. Through a technical and functional comparison with other market-leading tools such as Zapier and Make, it argues why n8n, especially its Community Edition, represents an ideal solution for software development companies seeking flexibility, control over their workflows, and technological sovereignty. It also explores its API integration capabilities, its modular architecture, and its native support for AI. It also presents real-world application cases in workflows such as automated onboarding, ETL, continuous testing, and omnichannel support. Finally, the importance of adopting a collaborative approach between automation and development is discussed, overcoming the false dilemma between "low-code vs. high-code" and proposing a synergy that fosters secure, scalable, and sustainable digital products.

Application of KNN algorithm for predicting celiac disease using clinical and serological variables

Daniel Levano — 2025-09-30

Celiac disease is an autoimmune condition with a global prevalence close to 1%, often underdiagnosed due to low clinical suspicion, which increases both morbidity and mortality. In this context, the application of the K-Nearest Neighbors (KNN) algorithm emerged as a predictive model to support the detection of this disease using clinical and serological variables. A supervised model was developed using the KNN algorithm and clinical and serological data extracted from an academic dataset containing 2,206 records. To address class imbalance, the Synthetic Minority Over-sampling Technique (SMOTE) was applied. The data were split for training and validation, optimizing the classification parameter through cross-validation. In addition, a web platform was developed to support data input, analysis, and output, allowing the uploading, processing, and generation of medical reports with role-based access and diagnostic probability estimation. The model achieved 94% accuracy, 97% precision, and 91% sensitivity. The algorithm proved to be effective for predicting celiac disease based on clinical and serological data, and its web-based implementation enables practical integration in clinical environments.

Cost Evaluation in Inventory Management: A Simulation Study Based on Python

Francisco Bencomo Murga — 2025-09-30

This article presents the results of a research project whose objective was to design a simulation process using a stochastic-dynamic model and the Jupyter Notebook tool with Phyton programming, to understand and analyze the behavior of the costs associated with an inventory management system with stochastic demand and periodic review, in a walnut distribution company. Data was collected on inventory holding costs, shortage costs, ordering costs, warehouse capacity, delivery times and demand for the last 6 months. The results obtained revealed that, over a 3-month horizon, the operating cost presents a parabolic behavior in relation to demand, which allows identifying the optimal or equilibrium point between inventory and total average costs, depending on an uncertain demand.

Web-based commercialization system to improve efficiency and logistics management at LAKTOMART S.A.

Juan Rodrigo Villanueva Ramos — 2025-09-30

The purpose of the research was the improvement of efficiency and logistics management in the area of commerce and distribution of the LAKTOMART S.A. company through the development of an online platform using the Model-View-Controller (MVC) architecture. A pre-experimental design was adopted with a quantitative orientation and had an explanatory scope. The methodology applied was the RUP (Rational Unified Process). For the implementation of the web application, HTML5, CSS3, JQuery V3.7.1, PHP V8.2.0 were used with the Laravel V10 framework. .48.180, domPDF V0.8 and MySQL V8.0.31 for database administration in addition, the evaluation of the economic indicators linked to the sales and distribution subsystems generated positive results that confirmed the economic viability of the developed system in question. The hypotheses were evaluated using the Shapiro-Wilk normality tests, and inferential statistical tests were applied, such as the student’s t test for related samples and the non-parametric Mann-Whitney test for independent samples, managing a significance level of. 5%. R statistical software was used for all tests. Based on the findings, it was concluded that the online platform with MVC architecture improves efficiency and effectiveness in LAKTOMART S.A.'s electronic business.

Deep Learning Techniques and Tools for Intelligent Weather Forecasting

Kevin Esteeven Parimango Gomez — 2025-09-30

In this paper, an analysis of deep learning techniques for weather forecasting using statistical downscaling approaches was developed. These are important, since they allow adjusting large-scale climate projections generated by the GCM climate model to more accurate and defined forecasts for specific areas, thus allowing overcoming the limitations of traditional numerical models in the representation of local and small-scale phenomena. Studies implementing Convolutional Neural Networks (CNN) and Generative Adversarial Networks (GAN) were analyzed in order to improve the spatial and temporal resolution of climate data. Both tools and techniques have proven to be effective in projects such as VALUE, which is in charge of evaluating downscaling methods in Europe, and DL4DS, a Python library in charge of applying deep learning algorithms to empirical downscaling of climate data. The main objective of this paper was to analyze the effectiveness of both tools and techniques focused on accuracy, scalability and computational efficiency, providing a complete overview of their use for the improvement of local weather forecasting.

A Case Study of the Salsa20 Encryption Algorithm Using Random Noisy Injection Enhanced by Artificial Intelligence

Edgar Rangel Lugo — 2025-09-30

Summary
The loss of digital data can have severe financial consequences for organisations. This situation can handle with a dynamic encryption approach, which it can generate multiple ciphertexts from a single plaintext. This paper delves into the intersection of artificial intelligence, random noisy schemes, and cybersecurity, analyzing several critical aspects of this emerging field. The application of random noisy scheme to ciphertext, it has been suggested as a means of enhancing cybersecurity, with recent studies indicating that this approach can be effective against cybercriminals. The objective of this research was focused on applying random noisy strategy on ciphertext of the Salsa20 encryption algorithm, as this area remains unexplored. Given the promising results of random noisy strategies, organisations that they employ Salsa20 may benefit from this approach. This research is important because recent experiments have revealed the effectiveness of using artificial intelligence for noisy injection on ciphertext. Therefore, is here presented a new opportunity for organisations regarding the employment of the novel random noisy Salsa20 encryption alternative because the experimental results have shown an increasing of the dynamic performance on ciphertext. This work introduces the random noisy Salsa20 strategy as a novel dynamic encryption alternative, and comparison of the performance to four random noisy schemes based on DES, 3DES, AES-256, and Blowfish algorithms. In conclusion, the novel alternative that it is here recommended, it can be difficult for the cybercriminals to decrypt.
Keywords: Applications of AI, cryptography, dynamic encryption methods, noisy injection strategies.

The Impact of Artificial Intelligence on the Evolution of Ad vanced Cities: An Empowered Future

Renzo Florian Villegas — 2025-09-30

This scientific study focuses on a literature review on the use of Artificial Intelligence (AI) and its impact on the different key sectors that make up a city. Highlighting the benefits of its implementation and the challenges faced by governments for its regulation and control. For this purpose, 9 scientific articles were selected that address the topics that will be discussed throughout this article: politics, transportation, citizen security, education, labor and health. The aim is to show the benefits of AI implementation in these sectors that will guide us towards a modern future where AI prevails as an irreplaceable tool.

Identification and Measurement of Self-Technical Debt in Deep Learning Frameworks: A Systematic Review

Elizabeth Cuatecontzi Cuahutle — 2025-09-30

Technical Debt in software development refers to the consequences of decisions prioritizing quick solutions over optimal ones. This concept, introduced by Ward Cunningham in 1992, has been widely studied to improve software quality. In the context of deep learning, Technical Debt is also present due to the use of tools that, while facilitating model creation, may generate debt and negatively impact performance.

Through a three-phase process, this study presents a systematic literature review to identify the types of Technical Debt found in deep learning tools and the techniques used for its identification and measurement. The reviewed studies show that Technical Debt can arise in various development phases, such as design, requirements definition, testing, documentation, source code, algorithms, and compatibility. Other affected aspects include data, models, knowledge, and infrastructure. Several approaches have been used to identify technical debt, such as analyzing comments in static code, pull requests, and commits, applying manual techniques, text mining, neural networks, and natural language processing algorithms. In terms of measurement, statistical methods are predominantly used.

The findings of this review provide a better understanding of how Technical Debt impacts deep learning tools and offer a foundation for guiding future research on its management and mitigation in the development of systems within intelligent environments.

La Innovations in Artificial Intelligence for Surgery Assistance: Systematic Review of Applications and Clinical Efficacy

Deysi Elvia Yuvixa Quiliche Plasencia — 2025-09-30

Artificial intelligence (AI) currently presents an innumerable number of implementations with widely proven results. Surgical assistance, a fundamental part of medical surgery, is a clear example of innovation and effective application of technologies that replicate different human qualities and skills to improve patient surgical care. The present research, through the PRISMA methodology, focuses on the analysis of AI applications for surgical assistance, making use of articles obtained after implementing a series of search, inclusion and exclusion criteria designed to achieve a total syntony between the scientific literature described and the central theme of the research. An in-depth search of the literature was performed in the ProQuest and Google Scholar databases and 18 articles were selected from a total of 272 candidates. The results show that technologies such as Machine Learning (ML), Deep Learning (DL), and Surgical Robotics are used to improve patient care with less invasive procedures and reduced risks. This study highlights the transformative impact of AI in surgical assistance and suggests the need for further research on its integration into new specialties, as well as ethical and regulatory aspects to ensure its safe use.

Systematic Review of the Debt Requirements Literature: Its Causes and Measurement

Maria Janai Sanchez Hernández — 2025-09-30

Technical Requirements Debt is defined as the difference between the initially stated requirements and the final software product. This study aimed to conduct a systematic literature review based on a methodology structured in three phases: definition of a search protocol, selection of relevant scientific sources, and application of inclusion and exclusion criteria, followed by a synthesis of the collected information. Research from the last five years was analyzed, considering a total of thirteen articles. The results indicate that the main causes of Technical Requirements Debt include the lack of formal documentation, pressure to meet deadlines, poor communication between the client and the development team, as well as the absence of automated tools that optimize requirements traceability, among other factors. Regarding its measurement, strategies such as cost-benefit analysis and rectification cost estimation have been proposed; however, these have not yet been validated in real-world contexts, which limits their practical applicability. In conclusion, Technical Requirements Debt represents a challenge in software engineering, directly affecting project quality and success. This work provides an updated overview that can serve as a basis for future research in the area, with the goal of developing more effective strategies for its management and possible mitigation.

Analyzing key attributes of AI and technologies for biometric authentication in smart hospitals

Carlos Daniel Gutiérrez Sandoval — 2025-09-30

This systematic review examines the use of artificial intelligence (AI) in biometric authentication for smart hospitals, with a particular focus on identifying the most efficient and widely used technological approaches in the world to improve the security and protection of medical data from unauthorized access. The integration of artificial intelligence (AI) through neural networks and machine learning algorithms with biometric security has been demonstrated to enhance the precision of individual identification and the detection of irregular behavior that may indicate unauthorized access. The PRISMA methodology entailed a comprehensive search of scientific studies employing key terms in conjunction with Boolean operators, followed by the selection of pertinent articles in accordance with pre-established inclusion and exclusion criteria. The results demonstrate that the incorporation of AI into biometric authentication systems enhances security in terms of controlled access, protection, and data security. The analyzed studies indicate that the deployment of multimodal biometrics and advanced algorithms not only improves the reliability of the process, but also reduces false positives, which is crucial in the management of sensitive data. The combination of various biometric features, such as facial recognition and physiological signal analysis, has proven to be effective even in medical settings

Machine learning algorithms for dementia prediction: A systematic review

Marco Lucas Guido Haro — 2025-09-30

The following work addresses the identification of algorithms used in machine learning for the early detection of dementia or degenerative cognitive impairment, currently one of the main clinical and socioeconomic challenges of this century. It indicates the most relevant machine learning algorithms that, with their high reliability and effectiveness, are gaining ground in a much more technological world. The methodology used corresponds to the PRISMA declaration standards, using highly demanding research repositories such as SCOPUS, SCIELO, IEEE XPLORE, SAGE JOURNAL, and GOOGLE SCHOLAR, finding 15 works that met all established criteria. The results of the review in these works found many comparisons by academic study. Among the most widely used models are Random Forest and SVM, which have shown accuracies above 85% in multiple studies. The conclusions affirm the relevance of Machine Learning as a technological tool in the detection of dementia and its varieties, indicating opportunities for future research, particularly in more specific case studies where the use of technology is essential to assist humans.