Muffakham Jah College of Engineering and Technology

Mental stress is a dangerous health issue with heart diseases, depression, and low productivity, and it should be identified as soon as possible. Conventional unimodal methods are insufficient to understand the complexity of stress that is expressed both physiologically and behaviorally. To do this, we present H-C3AT-G (Hierarchical Cross-Modal Contrastive Attention Transformer with Graph Fusion), a new end-to-end multimodal stress detection framework. The model uses contrastive pretraining to align the modalities, uses graph neural networks to fuse adaptive features, and uses a hierarchical cross-attention transformer to learn multi-scale dependencies. Moreover, Monte Carlo dropout improves predictive accuracy, whereas knowledge distillation reduces the model size to be used on a wearable. Tests on benchmark data demonstrate that H-C3AT-G performs better than CNN-BiLSTM, multimodal transformers, or GAT-based fusion, with an accuracy of 94.8

A finite element–based comparative assessment was conducted to evaluate deformation, stress, strain, and heat flux behavior of three commonly used gear pump materials under identical loading and boundary conditions. Three commonly used engineering materials—grey cast iron, aluminum alloy (AlSi10Mg), and stainless steel (316L)—were analyzed under varying pressure (1–3 MPa) and temperature (25–50 °C) conditions. Key performance metrics including total deformation, von Mises stress, elastic strain, and heat flux were examined to assess material suitability for high-pressure pump applications. Results revealed that stainless steel exhibited the lowest deformation (∼0.017 mm) and most uniform stress distribution under maximum load, while aluminum alloy demonstrated superior thermal conductivity but higher strain. Grey cast iron showed intermediate thermal performance but non-linear mechanical behavior indicative of brittleness. A 3D prototype was also fabricated using Fused Deposition Modeling (FDM) for physical visualization. The findings confirm stainless steel as the most structurally reliable material, making it the optimal choice for gear pumps operating under combined thermal and pressure stresses. The study is intended as a comparative simulation-based assessment supported by analytical benchmarking, rather than a fully experimentally validated predictive model.

Large language models hold considerable promise for various applications, but their computational requirements create a barrier that many institutions cannot overcome. A single session using a 70-billion-parameter model can cost around $127 in cloud computing fees, which puts these tools out of reach for organizations operating on limited budgets. We present AgentCompress, a framework that tackles this problem through task-aware dynamic compression. The idea comes from a simple observation: not all tasks require the same computational effort. Complex reasoning, for example, is far more demanding than text reformatting, yet conventional compression applies the same reduction to both. Our approach uses a lightweight neural controller that looks at the first few tokens of each request, estimates how complex the task will be, and sends it to an appropriately quantized version of the model. This routing step adds only about 12 milliseconds of overhead. We tested the framework on 290 multi-stage workflows from domains including computer science, physics, chemistry, and biology. The results show a 68.3% reduction in computational costs while preserving 96.2% of the original success rate. These findings suggest that routing queries intelligently can make powerful language models substantially more affordable without sacrificing output quality

The integration of artificial intelligence with embedded systems has significantly advanced the development of intelligent robotic platforms capable of human-like interaction. This paper presents the design and implementation of a real-time human mimicking robotic face system that utilizes computer vision and embedded control to replicate human facial movements. The system is developed using a Raspberry Pi-based architecture integrated with OpenCV for face detection and tracking, along with servo-controlled actuation for facial expressions. The proposed system focuses on achieving real-time responsiveness and synchronization between visual input and mechanical output. Experimental evaluation demonstrates an average face detection accuracy of approximately 92.4% under controlled lighting conditions, with a response latency of 180–250 milliseconds. The system also demonstrates stable performance in tracking and mimicking head movements within a defined range. This work contributes to the development of low-cost, intelligent robotic systems capable of human interaction, with potential applications in assistive robotics, human-computer interaction, and social robotics. Furthermore, the study highlights the importance of integrating security considerations into intelligent systems, as increasing system connectivity introduces potential vulnerabilities. Additional Keywords and Phrases : Facial actuation system, Jaw–Lip synchronization, Low-cost humanoid robotics, Raspberry Pi–based robot, Real-time speech interaction, large language model integration, Eye blink simulation

Pancreatic cancer remains one of the most lethal malignancies worldwide, with a five-year survival rate of less than 10%, making early and accurate diagnosis critical for improving patient outcomes. Histopathology continues to be the clinical gold standard, yet studies show that misclassification rates can range from 15% to 20% due to overlapping tissue structures and subjective interpretation by pathologists. Manual screening methods are often constrained by inter-observer variability and the complexity of distinguishing between benign pancreatic tissue and malignant regions, leading to diagnostic delays and potential inaccuracies. To address these limitations, this research proposes an innovative decision support named eXplainable Artificial Intelligence based Pancreatic Cancer Classification Network (XAI-PCC-Net). The methodology utilizes histopathology images from both Haematoxylin and Eosin (H&E) and May Grunwald-Giemsa (MGG) datasets, ensuring a diverse representation of tissue morphology. Gradient-weighted Class Activation Mapping Driven Dense201 Network (GCAM-D201-Net) is employed for feature extraction, offering a localized visualization of discriminative regions relevant to classification. These extracted features are then classified using a Deep Learning Important Features-Explainable Boosting (DLIFEB) classifier, which not only enhances predictive accuracy but also maintains interpretability for clinical decision support. The classification framework focuses on two primary categories: normal pancreatic tissue and pancreatic cancer tissue. By integrating explainability at both the feature extraction and classification stages, XAI-PCC-Net aims to provide reliable, transparent, and clinically viable diagnostic support to reduce human error and improve early pancreatic cancer detection.