Software development projects increasingly require project managers

who can interpret complex data, anticipate risks early, and make informed decisions

in rapidly changing Agile environments. Yet traditional project management

education—both university-based and vocational—often lacks realistic,

data-rich learning contexts where learners can analyze project behavior, explore

risk scenarios, and understand how estimation errors propagate. This paper presents an AI-based dashboard that supports early risk detection in Agile software

projects while simultaneously serving as a training platform in formal project

management (PM) education. The dashboard integrates ML models to predict

task duration deviations, detect sprint-level risks, and provide interpretable analytics.

Beyond real-time predictive insights, the dashboard functions as a learning

simulator for PM training programs. Students can upload or simulate project datasets, test planning assumptions, adjust risk thresholds, visualize correlations,

and observe the consequences of decision-making in a controlled environment,

The contributions of this work are: an interactive dashboard for risk-oriented project

analytics and a pedagogical model for integrating AI supported dashboards

into PM education.

Deep Neural Networks (DNNs) achieve strong performance

in semantic segmentation for robotic perception but remain vulnerable

to adversarial attacks, threatening safety-critical applications. While ro-

bustness has been studied for image classification, semantic segmenta-

tion in robotic contexts requires specialized architectures and detection

strategies.

We propose a framework for detecting adversarial attacks using pre-

trained ResNet-18 and ResNet-50 models. Our method leverages ad-

vanced feature extraction and statistical metrics to distinguish clean

from adversarial inputs. Experiments demonstrate its effectiveness across

various attacks, offering insights into model robustness. Additionally, we

compare network architectures to identify factors that enhance resilience.

This work supports the development of secure autonomous systems by

providing practical detection tools and guidance for selecting robust seg-

mentation models.

Children in today's world are used to studying through pictures, and the use of visual media in the classroom frequently improves their learning. This article describes a system that helps preschoolers learn the Latin alphabet's letters.

Given the advantages of using the idea of gamification in the classroom, an entertaining and captivating game was developed to assist students in drawing connections between letters and pictures of animals. Children can use the application, which is built on convolutional neural networks, to learn animal names more quickly by connecting dots on the screen to draw letters.

Through the use of different programming languages, the effective training of a letter recognition

model and the development of an intuitive graphical user interface, it has been demonstrated that a practical and attractive application has been created, which has been well received by children.

It has been shown that a useful and appealing application has been developed, which has been well-received by children, through the use of various programming languages, the efficient training of a letter recognition model, and the creation of an intuitive graphical user interface.

In order to make the application more kid-friendly and portable, a Raspberry BI was utilized together with a display in order to show letters generated from the application and an audio system for instructions.

This paper describes the change in teaching practice when delivering Medical Electronics course to the students from two faculties: students enrolled to Applied Electronics specialization from Faculty of Electrical Engineering and students enrolled to Biomedical Engineering from Faculty of Product Design and Environment at Transilvania University Romania. In previous years there was a lack of interaction between the groups of student from these two faculties so the teaching staff has decided to introduce the problem-oriented project work as summative assessment tool alongside the written exam in a safe learning environment which encouraged creativity. The activities included in the group projects followed the KASH model (Knowledge, Attitude, Skills, Habits) because employability represents a combination of knowledge, skills, attitudes, and personal attributes of graduates enabling them to contribute effectively to the workplace and society progress in general.

CRISPR/Cas-based biocomputation has enabled the design of advanced logic circuits analogous to CPUs. However, the scalability of these systems is limited by the lack of formal verification frameworks. Synthetic genetic circuits (SGCs) fail unpredictably due to resource competition (metabolic load), off-target effects, and crosstalk. The design-build-test cycle is slow and incapable of guaranteeing robustness. This article explores the use of Evolutionary (NEP) and Genetic (NGP) Processor Networks as a unified Bio-CAD (Biological Computer-Aided Design) framework. A conceptual mapping is presented where NGP processors model logic gates, strings model gRNAs and mRNA, and formal filters (rc_s, rc_w) model binding specificity. We argue that the inherently parallel and distributed architecture of NGP, and its demonstrated ability to solve NP-complete problems in linear time, directly reflects the massive parallelism of molecular computation in vivo. This approach would allow for the formal design and verification of fidelity, robustness, and scalability of biocomputers before their costly wetware synthesis.