Daily Archives: January 24, 2026

Authors: Jigarkumar Ambalal Patel, Mayur Girish Taunk

Abstract: The food industry is one of the largest global employers, yet it faces ongoing challenges in demand–supply chain management and food safety due to heavy reliance on manual processes and human error. Artificial Intelligence (AI) and Machine Learning (ML) are increasingly being adopted to transform the industry across the entire "farm to fork" pipeline by improving efficiency, accuracy, and safety. This paper reviews key AI- and ML-driven applications, including smart farming for crop monitoring and yield optimization, automated product sorting and grading, electronic noses for spoilage detection, and vision-based dietary assessment. Despite these advances, significant challenges remain, such as inaccurate image segmentation, high intra-class variation in food appearance, and the lack of large, standardized datasets. Overcoming these limitations is crucial for enabling reliable and scalable real-world deployment of AI.

DOI: http://doi.org/10.5281/zenodo.18359803

Published by: vikaspatanker

Authors: Jigarkumar Ambalal Patel, Mayur Girish Taunk

Abstract: The food industry is one of the largest global employers, yet it faces ongoing challenges in demand–supply chain management and food safety due to heavy reliance on manual processes and human error. Artificial Intelligence (AI) and Machine Learning (ML) are increasingly being adopted to transform the industry across the entire "farm to fork" pipeline by improving efficiency, accuracy, and safety. This paper reviews key AI- and ML-driven applications, including smart farming for crop monitoring and yield optimization, automated product sorting and grading, electronic noses for spoilage detection, and vision-based dietary assessment. Despite these advances, significant challenges remain, such as inaccurate image segmentation, high intra-class variation in food appearance, and the lack of large, standardized datasets. Overcoming these limitations is crucial for enabling reliable and scalable real-world deployment of AI.

DOI: http://doi.org/10.5281/zenodo.18359803

Published by: vikaspatanker

Authors: Jigarkumar Ambalal Patel, Mayur Girish Taunk

Abstract: The food industry is one of the largest global employers, yet it faces ongoing challenges in demand–supply chain management and food safety due to heavy reliance on manual processes and human error. Artificial Intelligence (AI) and Machine Learning (ML) are increasingly being adopted to transform the industry across the entire "farm to fork" pipeline by improving efficiency, accuracy, and safety. This paper reviews key AI- and ML-driven applications, including smart farming for crop monitoring and yield optimization, automated product sorting and grading, electronic noses for spoilage detection, and vision-based dietary assessment. Despite these advances, significant challenges remain, such as inaccurate image segmentation, high intra-class variation in food appearance, and the lack of large, standardized datasets. Overcoming these limitations is crucial for enabling reliable and scalable real-world deployment of AI.

DOI: http://doi.org/10.5281/zenodo.18359803

Published by: vikaspatanker

Authors: Mohammad Sameer Hussain, Jaspreet Kaur, Er. Gundeep Kaur

Abstract: Decision making systems are using a combination of style rules and new style artificial intelligence to help people make good choices. The old style rules are good because they are clear and easy to understand and they make sure people follow the rules. The old style rules have some problems though. They are hard to scale up. They cost a lot to maintain. Decision making systems that use style rules do not adapt well to new situations. On the hand the new style artificial intelligence like the kind that understands human language can find patterns and help with tough decisions. The style artificial intelligence is really good, at helping people make good choices because it can understand what people are saying and find patterns that the old style rules cannot. The style artificial intelligence is a big help to decision making systems because it can do things that the old style rules cannot. Decision making systems that use the style artificial intelligence can make better choices because they have more information and can understand what people are saying. This kind of intelligence has some problems. Artificial intelligence can make things up. It can be hard to understand intelligence. Also when something goes wrong with intelligence systems like these artificial intelligence systems it is not clear who is responsible, for the artificial intelligence. This paper reviews expert perspectives on both approaches and compares them in terms of interpretability, robustness, data dependence, deployment constraints, and evaluation. Evidence across multiple domains suggests that hybrid architectures integrating explicit rules, structured knowledge, and generative components provide a practical path toward trustworthy and adaptive decision- making.

DOI: http://doi.org/10.5281/zenodo.18359722

Published by: vikaspatanker

Authors: Leo. Tanyam. Encho, Abraham Okolo

Abstract: The Gompertz distribution is widely applied in describing human mortality, establishing actuarial tables, and various other fields. Historically, it was originally introduced by Benjamin Gompertz (1825) in connection with human mortality. This study aims to derive and analyze the mathematical and statistical properties of the Gompertz distribution, providing explicit expressions for parameter estimation from both frequentist and Bayesian perspectives. We then apply these estimation methodologies to analyze COVID-19 data in Cameroon. We investigate and compare numerous frequentist approaches for parameter estimation, including maximum likelihood, method of moments, pseudo-moments, modified moments, L-moments, percentile-based, least squares (including weighted), maximum product of spacings, minimum spacing absolute distance, minimum spacing absolute-log distance, Cramér-von-Mises, and Anderson-Darling (including right-tail) estimators. Their performance is evaluated using extensive numerical simulations, and their coverage probabilities are also assessed. Our results indicate that among the frequentist estimators, modified moments and moments estimators generally perform better than their counterparts. For Bayesian estimators, those based on the Mean Squared Error Loss Function (MSELF) and Kullback-Leibler Loss Function (KLF) demonstrate superior performance. The maximum product of spacings estimators also exhibit competitive performance.

DOI: https://doi.org/10.5281/zenodo.18359495

 

Published by: vikaspatanker

Authors: Ravi Teja Yarlagadda

Abstract: The increasing adoption of globally distributed cloud architectures has intensified the need for deployment mechanisms that ensure consistency, reliability, and scalability across multiple geographic regions. Traditional deployment approaches, often reliant on manual coordination or fragmented automation, struggle to meet the operational demands of modern cloud-native systems, leading to configuration drift, delayed releases, elevated failure rates, and prolonged recovery times. In this context, automated Continuous Integration and Continuous Deployment (CI/CD) pipelines integrated with Infrastructure-as-Code (IaC) have emerged as a promising paradigm for managing complex, multi-region cloud deployments; however, their systemic behavior, scalability characteristics, and reliability properties remain insufficiently explored at an empirical and analytical level. This study presents an in-depth evaluation of an automated CI/CD framework tightly coupled with Infrastructure-as-Code for multi-region cloud deployments, analyzed under medium-scale, production-like conditions. The research adopts a design-oriented experimental methodology to examine pipeline execution dynamics, failure semantics, resource utilization patterns, and recovery behavior across geographically distributed cloud regions. Infrastructure and application deployments are treated as deterministic, version-controlled artifacts, enabling systematic analysis of deployment repeatability, configuration convergence, and fault isolation. Comprehensive results demonstrate that the proposed CI/CD–IaC framework significantly enhances deployment performance and operational stability. Stage-wise analysis reveals low temporal variance and predictable execution behavior, while scalability experiments show sub-linear growth in deployment time as the number of target regions increases. Reliability metrics indicate consistently high availability exceeding 99.9%, with low mean time to recovery and strong isolation of regional failures. Failure characterization further confirms that most deployment anomalies are detected early in the pipeline lifecycle, minimizing downstream impact. Resource utilization analysis identifies build and testing stages as the dominant computational cost, validating the efficiency of state-aware infrastructure provisioning. Importantly, repeated deployments exhibit near-zero persistent configuration drift, confirming the system’s convergence toward a stable desired state. Overall, the findings establish that automated CI/CD pipelines integrated with Infrastructure-as-Code transform multi-region cloud deployment from a fragile, human-driven process into a resilient, self-stabilizing distributed system. This work contributes empirical evidence and system-level insights that advance understanding of deployment automation as a controlled and scalable engineering discipline, providing a foundation for future research in autonomous cloud operations, adaptive deployment pipelines, and AI-driven infrastructure management.

DOI: https://doi.org/10.5281/zenodo.18358912

 

Published by: vikaspatanker

Authors: Mr.Santosh Handignoor, Mr.Himanshu Singh, Prof. Vaishali Suryawanshi, Prof. Dipak Kadve

Abstract: Interview readiness is a decisive factor in determining employability and professional advancement; however, a large number of students struggle to perform effectively due to limited practice opportunities, anxiety, and the absence of structured, objective feedback. Recent developments in Artificial Intelligence (AI) have enabled the creation of intelligent systems capable of simulating interview scenarios and evaluating candidates in a consistent and data-driven manner. This research examines an AI-based mock interview framework that utilizes Natural Language Processing for response evaluation, speech analytics for assessing confidence and fluency, and facial expression analysis for understanding non-verbal behavior. By combining these AI techniques, the system delivers personalized feedback that highlights communication gaps, confidence issues, and knowledge deficiencies. Unlike traditional mock interviews, the proposed approach allows repeated practice without dependency on human evaluators, ensuring scalability and fairness. The study is supported by quantitative analysis conducted on a student dataset, revealing notable improvements in interview performance, self-confidence, and communication effectiveness after exposure to AI-driven mock interviews. The results demonstrate that AI-based interview preparation tools can significantly enhance interview readiness and serve as an effective alternative to conventional training methods. This work reinforces the growing role of AI in employability skill development and its potential to transform interview preparation practices in academic and recruitment environments.

DOI: https://doi.org/10.5281/zenodo.18387039

 

Published by: vikaspatanker