High-density 3D photonic integration for scalable laser neural networks

Cloud workload evolution has progressed because end-users need real-time applications such as autonomous systems, industrial IoT, and innovative healthcare. Traditional cloud computing systems cause substantial latency because they process information centrally while sending and receiving data. Artificial intelligence and Edge computing unite to provide an effective solution through network edge-based computations distribution, enabling fast real-time data processing. This research analyzes important strategies used in edge computing and artificial intelligence technology to minimize delays in cloud computing operations. The paper introduces basic principles of edge computing with AI capabilities for cloud workload management. We examine three main challenges: network bottleneck, processing capacity limitations, and security threat considerations. The proposed solution incorporates edge AI accelerators with hardware implementation and lightweight AI models, federated learning and reinforcement learning as software approaches, and 5G technology and edge caching as network optimization methods. The document demonstrates real-world applications based on case studies and experimental outcomes from autonomous vehicles, the healthcare sector, and industrial IoT implementations. The paper conducts a systematic evaluation that compares edge-based AI systems against traditional cloud computing based on their latency, power efficiency, and flexibility performance. It examines upcoming technologies that involve AI-driven self-adjusting edge networks and combinations of cloud and edge AI platforms. The discussion grows in technical depth through the addition of graphs, flowcharts, pseudocode, and system diagrams. Our research confirms that AI integration with edge computing lowers end-to-end latency while improving real-time decisions and maximizing cloud-based resource performance. The study establishes a framework for developing next-generation cloud architectural infrastructure that uses low-latency artificial intelligence.

Coherent optical communication is considered as an indispensable solution to the ever-increasing demand for higher data rates. To reduce the cost and form factor of coherent transceivers, full integration of photonic devices including lasers, modulators, amplifiers, photodetectors, and other components is necessary. However, as fabricating optical isolators on chip remains extremely challenging, optical feedback, which can degrade the coherence of semiconductor lasers, becomes the main obstacle, thwarting large-scale photonic integration. An appealing solution to such a problem is to use semiconductor lasers with intrinsic insensitivity to optical feedback as the integrated light sources. The heterogenous Si/III-V lasers, with their built-in high-Q resonators, are expected to possess a robustness to optical feedback which exceeds by several orders of magnitude compared to commercial III-V distributed feedback (DFB) lasers, which will be validated here. We present data showing that the heterogeneous Si/III-V lasers can preserve their phase coherence under much larger optical feedback and therefore function without severe degradation in isolator-free coherent optical communication systems.

InP based optical and electronic devices have been extensively studied for their enabling properties in modern telecommunication systems. There is considerable demand to further improve the performance of lasers and in particular to integrate the lasers with additional optical components such as detectors, waveguides, amplifiers, splitters, interferometers, modulators and multiplexers to realize optical chips with advanced functionality. This is hindered by the dominating costs associated with integrating these devices. On-chip integration is an attractive solution but one which requires the use of complex technologies (electron beam lithography and epitaxial overgrowth) that are generally proprietary and not compatible with foundry manufacture. In this talk, we report on a novel alternative approach to integration using a single epitaxial growth stage and conventional optical lithography. A tuneable laser with an integrated optical amplifier is demonstrated but the technology can be extended to integration of modulators and other optical functions. This device is enabled by the realization of a reflective mirror using etched slots thus freeing the laser designer from the absolute requirement of the cleaved facet and the associated crystal orientation of all semiconductor lasers. The integration of multiple components on a single chip will provide a platform for photonics that can be implemented in a foundry and this mimics the approach of the silicon electronics industry.

Integrated indium phosphide (InP) based semiconductor laser systems with extended cavities can be realized using photonic integration technologies that combine optical amplifiers with devices based on passive ridge waveguides with a quaternary InGaAsP waveguide core [1] or pure InP nanophotonic waveguides (IMOS technology) [2] . Extended cavities can be utilized for lasers with tunable filters or mode locked lasers (MLL) with long cavities and fine-tuning elements [3] . In these passive waveguide sections two-photon absorption (TPA) can occur, which in turn gives rise to free carrier absorption (FCA) from high energy carriers and a change in refractive index. In addition to that the refractive index of the material is intensity dependent. Observations in MLLs with intra-cavity phase modulators [3] and of linewidths of the modes that are much wider than those reported for extended cavity MLLs using silicon waveguides [4] , may be due to TPA generated carriers.

This paper explores the integration of generative AI and large language models into the realm of software engineering education and training, with a specific focus on the transformation of traditional peer assessment methodologies. The motivation stems from the growing demand for innovative educational techniques that can effectively engage and empower learners in mastering Software Engineering principles. The proposed approach involves presenting students with modeling exercises solved by ChatGPT, prompting them to critically evaluate and provide constructive feedback on the generated solutions. By engaging students in a dialogue with the AI model, we aim to foster a dynamic learning environment where learners can articulate their considerations and insights, thereby enhancing their comprehension of software engineering principles, critical thinking and self evaluation skills. Preliminary results from pilot implementations indicate promising outcomes, suggesting that this approach not only enhances the quality of peer feedback but also contributes to a more interactive and engaging educational experience.

L2 memory, serving multiple clusters of tightly coupled processors, is well-suited for 3D integration, given its large required size and tolerance to latency and variations in memory access time. In this paper, we focus on the design of a synthesizable L2 memory IP component, which can be attached to a cluster-based multi-core platform through its NoC ports, and offer high-bandwidth memory access with low average latency. We propose a scalable 3D nonuniform memory access (NUMA) architecture, based on low latency logarithmic interconnects, which allows stacking of multiple memory layers with identical dies, supports multiple outstanding transactions, and achieves high clock frequencies due to its highly pipelined nature. Benchmark simulation results demonstrate that addition of 3D-NUMA to a multi-core NoC can result in an average performance boost of 34%. Physical synthesis results show that 3D-NUMA memory system can operate at 500 MHz in STMicroelec-tronics CMOS-28nm Low Power Technology (bounded by memory cut access time, while its logic components can operate up to 1 GHz), up to 8 layers (4 MB) with a memory density loss of only 16%.

• Significant transparency and responsible modelling are vital for high-stakes AI forecasting. • The reported forecast accuracy of Google’s AI flood model is substantially overstated when evaluated against real observational data. • Google’s AI flood model relies on several subjective and questionable evaluation choices. The development of AI models is increasing at a rapid rate. However, when are they ready to be deployed in real-world operational settings? In this paper, we introduce a framework to support such assessments and apply it to Google’s recently released AI-based flood prediction system, which is claimed to achieve “reliability in predicting extreme riverine events” and provide “accurate and timely warnings” that are available “earlier and over larger and more impactful events in ungauged basins”. The system has been integrated into an operational early-warning platform producing open, real-time forecasts in more than 80 countries. While this development promises to usher in a new and exciting age in global flood forecasting, the supporting evidence relies heavily on several subjective choices, the implications of which have not been acknowledged or assessed. Here, we evaluate the consequences of these choices on claims of operational deployment readiness across four dimensions: predictive accuracy, forecast timeliness, the characterization of extreme events, and benchmarking against state-of-the-art models. Our assessment reveals that the system’s actual predictive accuracy is likely to be substantially lower than reported—particularly for extreme events—raising concerns about responsible practices across modelling and publicity in high-stakes applications. The deployment of the Google AI model therefore risks misinforming those who depend on its outputs for evacuation and preparedness decisions, particularly in less-developed countries such as those targeted by the enterprise, given its alarmingly high (>90%) rates of false positives and false negatives. Beyond the immediate operational consequences, if left unaddressed, these outcomes may erode public trust in AI within hydrological sciences. We conclude by calling for greater transparency, accountability, and methodological rigor in the integration of AI into flood forecasting.

INTRODUCTION: With the rapid development of cloud computing and intelligent technology, traditional accounting informatization systems are gradually migrating to cloud computing platforms, promoting financial management's automation and intelligence. OBJECTIVES: However, data security in the cloud computing environment, especially the issue of data integrity, has become one of the core challenges for intelligent accounting informatization systems. Accounting informatization systems rely on the storage and processing of a large amount of financial, tax, and transaction data, and the integrity of these data is directly related to the financial transparency, compliance, and decision-making accuracy of an enterprise. METHODS: Therefore, how to design an efficient and scalable cloud data integrity verification algorithm becomes the key to ensuring the reliability and security of intelligent accounting systems. This paper proposes a cloud data integrity verification algorithm for smart accounting informationization based on a cloud computing environment. The algorithm combines the efficient data processing capability of cloud computing with the automation requirements of smart accounting systems to ensure data integrity during storage, transmission, and processing through various data validation techniques such as digital signatures, hash algorithms, and blockchain-based distributed ledgers. RESULTS: The algorithm design considers the dynamics and scalability of the cloud computing environment. It can realize fast, real-time data validation and integrity detection in a large-scale data environment. Through experimental verification, the algorithm proposed in this paper performs well in processing large-scale accounting data with high verification efficiency and accuracy. CONCLUSION: Compared with traditional validation methods, the algorithm improves the speed of data validation. It enhances the system's ability to prevent data tampering and loss, providing a new cloud data integrity assurance scheme for intelligent accounting informatization systems.

This article examines the transformative convergence of edge computing and artificial intelligence technologies, which is fundamentally reshaping infrastructure paradigms across industries. As computational intelligence moves closer to data sources, new architectures are emerging that address the critical requirements of latency-sensitive applications, data privacy concerns, and bandwidth optimization. The article explores the technological foundations enabling AI at the edge, including 1lightweight containerization, specialized hardware innovations, and energy-efficient computing approaches. The analysis extends to orchestration challenges in geographically distributed environments and the revolutionary potential of federated learning for privacy-preserving distributed intelligence. Through examination of real-world implementations across healthcare, manufacturing, and smart city contexts, the article identifies key performance metrics, optimization strategies, and lessons learned from early adopters. The discussion concludes with an assessment of emerging trends, research gaps, and standardization efforts shaping the future of edge-AI integration. This comprehensive overview provides Cloud Engineering professionals with essential insights for designing, deploying, and managing the next generation of intelligent distributed applications in an increasingly edge-centric computational landscape.

The growing adoption of large language models in healthcare has highlighted persistent challenges related to hallucinations, lack of traceability, and limited clinical trust. Retrieval-Augmented Generation (RAG) has emerged as a critical architectural advancement that integrates generative models with curated knowledge sources to support real-time, evidence-based clinical decision-making. This chapter conceptually examines how RAG systems enhance knowledge intelligence in healthcare by dynamically retrieving relevant medical literature, clinical guidelines, and patient-specific data before generating responses. By grounding outputs in verified sources, RAG architectures improve diagnostic reasoning, treatment recommendations, and clinical documentation while maintaining transparency and auditability. The chapter synthesizes recent developments in clinical decision support systems, knowledge graphs, vector databases, and retrieval pipelines to demonstrate how RAG reduces hallucinations and aligns AI outputs with medical evidence. It further explores RAG-enabled clinician workflows, including diagnostic support, drug–drug interaction checks, guideline adherence, and real-time summarization of electronic health records. Ethical, technical, and operational considerations such as data governance, latency, interoperability, and clinician oversight are also addressed. The chapter contributes a structured conceptual framework for deploying RAG-based clinical intelligence systems that balance automation with human expertise, offering insights for researchers, healthcare providers, and policymakers seeking safe and scalable AI integration in clinical environments.

The growing adoption of large language models in healthcare has highlighted persistent challenges related to hallucinations, lack of traceability, and limited clinical trust. Retrieval-Augmented Generation (RAG) has emerged as a critical architectural advancement that integrates generative models with curated knowledge sources to support real-time, evidence-based clinical decision-making. This chapter conceptually examines how RAG systems enhance knowledge intelligence in healthcare by dynamically retrieving relevant medical literature, clinical guidelines, and patient-specific data before generating responses. By grounding outputs in verified sources, RAG architectures improve diagnostic reasoning, treatment recommendations, and clinical documentation while maintaining transparency and auditability. The chapter synthesizes recent developments in clinical decision support systems, knowledge graphs, vector databases, and retrieval pipelines to demonstrate how RAG reduces hallucinations and aligns AI outputs with medical evidence. It further explores RAG-enabled clinician workflows, including diagnostic support, drug–drug interaction checks, guideline adherence, and real-time summarization of electronic health records. Ethical, technical, and operational considerations such as data governance, latency, interoperability, and clinician oversight are also addressed. The chapter contributes a structured conceptual framework for deploying RAG-based clinical intelligence systems that balance automation with human expertise, offering insights for researchers, healthcare providers, and policymakers seeking safe and scalable AI integration in clinical environments.

We experimentally show vertically stacked, multi-layer, low-temperature deposited photonics for integration on processed microelectronics. Waveguides, microrings, and crossings are fabricated out of 400°C PECVD Si3N4 and SiO2 in a two layer configuration. Waveguide losses of ~1 dB/cm in the L-band are demonstrated using standard processing and without post-deposition annealing, along with vertically separated intersections showing -0.04 ± 0.002 dB/cross. Finally 3D drop rings are shown with 25 GHz channels and 24 dB extinction ratio.

Optical technology plays an increasingly important role in numerous information system applications, including optical communications, storage, signal processing, biology, medicine, and sensing. As optical technology develops, there is a growing need to develop scalable and reliable photonic integration technologies. These include the development of passive and active optical components that can be integrated into functional optical circuits and systems, including filters, electrically or optically controlled switching fabrics, optical sources, detectors, amplifiers, etc. We explore the unique capabilities and advantages of nanotechnology in developing next generation integrated photonic information systems. Our approach includes design, modeling and simulations of selected components and devices, their nanofabrication, followed by validation via characterization and testing of the fabricated devices. The latter exploits our recently constructed near field complex amplitude imaging tool. The understanding of near field interactions in nanophotonic devices and systems is a crucial step as these interactions provide a variety of functionalities useful for optical systems integration. Furthermore, near-field optical devices facilitate miniaturization, and simultaneously enhance multifunctionality, greatly increasing the functional complexity per unit volume of the photonic system. Since the optical properties of near-field materials are controlled by the geometry, there is flexibility in the choice of constituent materials, facilitating the implementation of a wide range of devices using compatible materials for ease of fabrication and integration.

The wavelength division demultiplexer (WDD) is an indispensable component for separating and routing photons to spatially distinct detection nodes in practical quantum key distribution (QKD) systems. In this work, we propose an ultra-low-loss and broadband WDD based on lithium niobate on insulator (LNOI), designed and optimized through comprehensive three-dimensional simulations, utilizing subwavelength grating (SWG)-assisted double directional coupling for the separation of 780 and 1550 nm wavelengths. The SWG waveguide enables both a compact footprint and low insertion loss (IL) over a broad wavelength range, making it well suited for QKD applications. The optimized device, with a total length of 45 µm, achieves ILs of 0.04 dB at 780 nm and 0.22 dB at 1550 nm, along with extinction ratios of 22 and 32 dB, respectively. The 0.5 dB bandwidth exceeds 140 nm for 1550 nm and 230 nm for 780 nm. Furthermore, the structure exhibits strong tolerance to fabrication-induced variations, supporting its suitability for robust and scalable photonic integration.

Coronary Artery Disease (CAD) remains a leading cause of morbidity and mortality worldwide. To evaluate the diagnostic performance of AI models in the early detection and risk stratification of CAD using ECG and imaging data and to compare their accuracy with cardiologist interpretation. Methods: This cross-sectional analytical study was conducted in Saudia, included 185 patients with suspected CAD who underwent both ECG and imaging evaluation. AI-based models, including convolutional neural networks for ECG and deep learning algorithms for imaging, were applied to detect CAD and stratify patients into risk categories. Cardiologist-confirmed diagnosis served as the reference standard. Results: The AI-ECG model achieved a sensitivity of 88.5%, specificity of 82.0% and an AUC of 0.90 (95% CI: 0.86–0.94). When ECG and imaging data were combined, diagnostic accuracy improved, with sensitivity of 92.4%, specificity of 85.2% and an AUC of 0.93 (95% CI: 0.89–0.96). AI-based risk stratification categorized 54 patients (29.2%) as low risk, 78 (42.2%) as intermediate risk and 53 (28.6%) as high risk. Confirmed CAD prevalence correlated strongly with AI-predicted risk groups, with 22.2% in the low-risk group and 88.7% in the high-risk group. Conclusion: Artificial intelligence demonstrates high accuracy in the early detection and risk stratification of CAD using ECG and imaging data. AI models performed comparably to cardiologists and offered significant efficiency gains. Integration of AI into cardiovascular workflows may enable earlier intervention, optimized resource allocation and improved patient outcomes. Further validation across larger and more diverse populations is warranted.