International Journal of Science, Engineering and TechnologyA Systematic Review Of Deep Learning Approaches For Agricultural Monitoring Using Satellite Imagery
Authors: Vandana Birle, Dr. Dilip Singh Solanki
Abstract: The increasing demand for sustainable agricultural production has accelerated the adoption of advanced technologies for crop monitoring and farm management. Satellite imagery has emerged as a valuable source of large-scale and continuous agricultural data, while deep learning techniques have demonstrated remarkable capabilities in extracting meaningful information from complex remote sensing datasets. This systematic review examines recent advances in deep learning approaches applied to satellite imagery for agricultural monitoring. The review analyzes studies published between 2023 and 2026 focusing on crop classification, farmland segmentation, crop health assessment, disease detection, yield prediction, land-use mapping, and environmental stress monitoring. Various deep learning architectures, including Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), Transformers, Generative Adversarial Networks (GANs), and hybrid models, are discussed in relation to their effectiveness in agricultural applications. The findings indicate that deep learning significantly improves the accuracy and efficiency of agricultural monitoring compared to conventional image processing methods. However, challenges related to data availability, model generalization, computational complexity, and real-time implementation remain significant barriers to widespread adoption. The review highlights emerging trends and identifies future research directions for developing intelligent, scalable, and sustainable agricultural monitoring systems through the integration of satellite imagery and artificial intelligence.
DOI: http://doi.org/10.5281/zenodo.21412141
Finite Element Analysis Based Design and Optimization of a Lightweight Electric Vehicle Chassis
Authors: Research Scholar Abhinav Singh Dangi, Associate Professor Dr. Arun Kumar Yadav
Abstract: The increasing demand for energy-efficient and sustainable transportation has accelerated the development of lightweight electric vehicles. Vehicle weight plays a significant role in determining battery performance, driving range, energy consumption and overall vehicle efficiency. This study focuses on the lightweight design and optimisation of an electric vehicle chassis using advanced structural analysis and lightweight material strategies. The research reviews the application of lightweight materials such as aluminium alloys, magnesium alloys, high-strength steel and composite materials for reducing vehicle mass while maintaining structural strength and safety. Finite Element Analysis (FEA) and topology optimisation techniques are studied to improve chassis stiffness, load distribution and durability, and the relationship between lightweighting, energy efficiency, sustainability and intelligent chassis technologies is examined. Twelve studies published between 1998 and 2026 are reviewed and organised into three themes covering lightweight materials and design strategies, chassis optimisation and structural analysis, and energy efficiency with intelligent chassis performance. The reviewed evidence is consolidated into comparative tables that map each study to its focus, method, principal finding and limitation, alongside a qualitative comparison of candidate lightweight materials and a summary of the structural analysis methods reported. Five research gaps are identified and mapped to corresponding research directions. The findings indicate that optimised lightweight chassis structures can reduce vehicle weight, improve energy efficiency, enhance battery performance and increase driving stability, contributing toward the development of efficient, safe and sustainable electric vehicle systems.
Efficient Hybrid CNN-Transformer Models for Benchmark-Driven Image Deepfake Detection and Analysis
Authors: Research Scholar Priya Mishra, Dr. Deepika Pathak
Abstract: The rapid evolution of deepfake generation technologies has created significant challenges for digital media authentication, cybersecurity and misinformation prevention. Modern deepfake images generated using advanced artificial intelligence techniques exhibit highly realistic visual characteristics, making manual identification increasingly difficult. This paper presents a comprehensive review of efficient hybrid CNN-transformer models for benchmark-driven image deepfake detection and analysis. The study critically examines recent advances in Convolutional Neural Networks (CNNs), transformer-based architectures, hybrid feature fusion frameworks and explainable AI techniques for identifying manipulated images, and analyses benchmark-driven evaluation strategies, multimodal feature representation, manipulation localisation and attention-based contextual learning. Twelve studies published in 2025 are reviewed and organised into three themes covering hybrid architectures, benchmark-driven feature analysis, and cybersecurity challenges. The reviewed evidence is consolidated into comparative tables that map each study to its focus, approach, principal finding and limitation, alongside a component-level comparison of CNN, transformer and hybrid designs and an assessment of modality coverage across the corpus. Major challenges including adversarial attacks, computational complexity, cross-domain generalisation, multimodal deepfakes and benchmark inconsistency are discussed, and seven research gaps are identified and mapped to corresponding future research directions focusing on explainable AI integration, multimodal fusion, computational optimisation and generalised feature learning.
Gold Price Forecasting for Investment Decisions: A Multi-Factor Predictive Analytics Approach
Authors: Research Scholar Govind Kumar Verma, Associate Professor Dr. Richa Pareek
Abstract: Gold has long been considered one of the most important investment assets and safe-haven instruments in global financial markets. Rising market volatility, inflationary pressure, geopolitical uncertainty and shifting economic conditions have made accurate gold price forecasting essential for investors, financial institutions and policymakers. This study proposes a multi-factor predictive analytics approach to gold price forecasting that supports investment decision-making and risk management. The framework integrates machine learning, deep learning, econometric modelling and managerial analytics techniques to analyse the impact of macroeconomic indicators such as interest rates, exchange rates, inflation, commodity prices, climate risks and market volatility on gold price movements. Twelve studies published between 2017 and 2026 are reviewed and organised into three themes covering deep learning approaches, econometric and multi-factor models, and data mining with investment decision support. The reviewed evidence is consolidated into comparative tables that map each study to its data context, method, principal finding and contribution to the proposed design, and seven research gaps are identified and mapped to corresponding research directions. The framework is specified through five macroeconomic factor blocks and five forecasting components including Long Short-Term Memory (LSTM) networks, hybrid neural networks, ARIMA, Support Vector Machines and adaptive learning mechanisms. The findings indicate that multi-factor predictive analytics frameworks can improve forecasting accuracy, support strategic investment planning and enhance investor decision-making in uncertain financial environments.
Corporate Valuation Based on Dividend Policy in the FMCG Sector: A Review of Predictive Analytics Approaches
Authors: Research Scholar Jyoti Wadhwani, Associate Professor Dr. Uttam Kumar Jha
Abstract: Corporate valuation is central to financial management and investment analysis, and it is especially consequential in the Fast-Moving Consumer Goods (FMCG) sector, where stable demand, recurring cash flows and sustained investor confidence make firm value highly sensitive to distribution decisions. Dividend policy is repeatedly identified in the literature as a determinant of shareholder wealth, stock price behaviour, profitability signalling and investment attractiveness, yet the evidence remains fragmented across variables, markets and estimation methods. This paper presents a structured review of contemporary research on the dividend policy-corporate valuation relationship and on the emerging use of predictive analytics in valuation modelling. Twelve recent primary studies published between 2025 and 2026 were identified, screened and classified into four themes: dividend policy and corporate financial performance; dividend policy, stock market behaviour and investor response; capital structure, ownership and governance effects; and ESG, macroeconomic and analytics-driven valuation. Each study is compared on context, the structural role assigned to dividend policy, reported findings and methodological limitations. The synthesis shows that dividend policy operates in three distinct structural roles across the literature-as a direct determinant, as a moderator and as a mediator-and that this inconsistency is a substantive source of conflicting results. It further shows that the reviewed studies rely almost exclusively on conventional regression and structural equation techniques, that ESG and macroeconomic variables are treated in isolation, and that the FMCG sector itself is under-examined despite its dividend-intensive profile. Drawing on these findings, the review derives seven research gaps and proposes an integrated conceptual framework in which financial performance, capital structure, ownership, market and ESG determinants feed a dividend policy core, which in turn feeds a predictive analytics layer combining regression, machine learning, time-series forecasting and explainable AI to generate valuation forecasts and decision-support outputs. The framework is offered as a research agenda rather than an estimated model, and the review concludes with prioritised directions for its empirical validation.
Hybrid Deep Learning Framework for Robust and Computationally Efficient Image-Based Deepfake Detection
Authors: Research Scholar Priya Mishra, Dr. Deepika Pathak
Abstract: The rapid advancement of artificial intelligence and generative models has significantly increased the creation and dissemination of deepfake images, posing severe threats to digital security, media authenticity and public trust. Conventional deepfake detection methods often suffer from limited generalisation capability, high computational complexity and reduced performance against advanced synthetic image generation techniques. This review paper presents a comprehensive analysis of hybrid deep learning frameworks for robust and computationally efficient image-based deepfake detection. The study critically examines existing architectures including Convolutional Neural Networks (CNNs), Vision Transformers (ViTs), attention-based models and explainable artificial intelligence (XAI)-enabled frameworks, and investigates the role of preprocessing and feature enhancement techniques such as normalisation, edge detection, artifact analysis and feature fusion in improving detection accuracy. Eleven studies are reviewed and organised into three themes covering deep learning architectures, preprocessing and feature enhancement, and benchmarking with societal impact. The reviewed evidence is consolidated into comparative tables that map each study to its focus, approach, principal finding and limitation, alongside a comparison of architecture families and a summary of preprocessing techniques. The review identifies six research gaps related to computational efficiency, cross-dataset generalisation, model interpretability, preprocessing strategy, benchmark inconsistency and lightweight hybrid design, and maps each to a corresponding future research direction. Directions focusing on lightweight hybrid architectures, explainable frameworks, benchmark-driven evaluation and multimodal detection strategies are presented to support the development of reliable and scalable deepfake detection systems.
Two-Way Fluid-Structure Interaction Analysis of Composite Wind Turbine Blades for Improved Fatigue Resistance and Structural Durability
Authors: Research Scholar Rishabh Dev Singh, Associate Professor Dr. Banarsi Pandey
Abstract: The increasing demand for renewable energy has accelerated the development of large-scale wind turbine systems with improved aerodynamic efficiency, structural reliability and operational durability. Wind turbine blades are continuously subjected to complex aerodynamic loading, cyclic stresses, vibration and environmental conditions that significantly affect their structural integrity and fatigue performance. This study presents a two-way Fluid-Structure Interaction (FSI) analysis of composite wind turbine blades for improved fatigue resistance and structural durability. Advanced composite materials are used to design lightweight blade structures with enhanced stiffness and load-bearing capability, and two-way FSI simulations investigate the interaction between aerodynamic airflow and structural deformation under varying wind speed conditions. Structural analysis and fatigue simulations evaluate stress distribution, deformation characteristics, vibration response and fatigue life of the composite blade system. Twelve studies published between 2003 and 2021 are reviewed and organised into three themes covering blade design with aerodynamic optimisation, composite materials with structural analysis, and blade damage with monitoring and sustainability. The reviewed evidence is consolidated into comparative tables that map each study to its focus, method, principal finding and limitation, alongside an assessment of the coupling approaches the corpus employs and a summary of the load cases and failure drivers it addresses. Five research gaps are identified and mapped to corresponding research directions. The simulation results demonstrate that optimised composite blade configurations significantly improve aerodynamic stability, reduce stress concentration and enhance fatigue resistance relative to conventional designs, confirming that integrated two-way FSI analysis combined with advanced composite materials provides an effective computational approach for improving structural durability and long-term operational performance.
Aerodynamic Performance Evaluation of CFRP-Based Blended Winglets for Commercial Aircraft Using ANSYS Simulation
Authors: Satyaprakash Tiwari, Dr Alok Choudhary
Abstract: The increasing demand for fuel-efficient and environmentally sustainable commercial aircraft has encouraged the development of advanced aerodynamic technologies for drag reduction and performance enhancement. Winglets are widely used in modern aircraft to reduce induced drag, minimise wake turbulence and improve aerodynamic efficiency. This study focuses on the aerodynamic performance evaluation of Carbon Fibre Reinforced Polymer (CFRP)-based blended winglets for commercial aircraft using ANSYS simulation. A lightweight blended winglet model is developed using CFRP and glass-carbon composite materials to improve structural strength while reducing overall wingtip weight, and Computational Fluid Dynamics (CFD) analysis is performed to evaluate airflow characteristics, pressure distribution, lift generation, drag reduction and wake turbulence behaviour around the winglet structure. The performance of the optimised blended winglet is compared with a conventional wingtip configuration on aerodynamic efficiency, stability and drag characteristics. Twelve sources published between 2004 and 2023 are reviewed and organised into three themes covering commercial aircraft wing design and aerodynamic performance, composite materials and lightweight structures, and numerical simulation with safety and performance assessment. The reviewed evidence is consolidated into comparative tables that map each source to its focus, method, principal contribution and limitation, alongside an assessment of the disciplinary composition of the corpus and the aerodynamic evaluation parameters it addresses. Five research gaps are identified and mapped to corresponding research directions. The simulation results demonstrate that the CFRP-based blended winglet significantly reduces induced drag and wake turbulence while improving lift-to-drag ratio and overall flight performance.
Soft Sensor-Assisted Nonlinear Model Predictive Control for Pyrolysis Reactor Temperature Regulation
Authors: Ranjan Kumar
Abstract: Pyrolysis reactors are difficult to control because thermochemical decomposition is nonlinear, the reaction rates are high, and feedstock variability causes disturbances. Traditional proportional-integral-derivative (PID) controllers often perform poorly, causing overshoot, sluggish response, and suboptimal energy efficiency. To address these challenges, this study proposes a soft-sensor-enabled nonlinear model predictive control (NMPC) strategy for temperature regulation in a pyrolysis reactor. This method combines an Extended Kalman Filter (EKF) soft sensor with predictive control to estimate unmeasured disturbances and states. A nonlinear energy balance model considers the heat input, heat loss, reaction dynamics, and disturbances. The NMPC controller solves an optimization problem over the prediction horizon while satisfying the constraints on the control inputs and reactor temperature. Simulations show that it outperforms PID and traditional MPC controllers, achieving a 499.75°C steady-state temperature, 0.247°C tracking error, 0.014°C overshoot, and minimized RMSE and IAE values of 129.53°C and 68060, respectively. The EKF soft sensor achieved a 0.40°C RMSE, enabling disturbance compensation. This method provides comparable energy efficiency while improving stability and robustness. In conclusion, this strategy offers an accurate and efficient solution for nonlinear pyrolysis reactor temperature control under disturbance conditions.
Cyber Crime and Indian Legal Framework: An Analysis of Emerging Challenges and Necessary Reforms
Authors: Vithalanai Nirav Prakashbhai
Abstract: The rapid expansion of information and communication technologies has fundamentally transformed the manner in which individuals, businesses, governmental institutions and societies communicate, transact and store information. Alongside the benefits of digitalisation, cyberspace has become an increasingly significant environment for criminal activities such as identity theft, phishing, online financial fraud, cyber stalking, ransomware, data breaches, cyber extortion, online impersonation, cyber terrorism, deepfakes and AI-enabled crimes. The borderless and technologically dynamic nature of cybercrime creates significant challenges for conventional criminal law, investigation, evidence collection and prosecution. India has developed a multi-layered legal and institutional framework to address these challenges. The Information Technology Act, 2000 remains the principal legislation dealing specifically with cyber offences and electronic transactions. However, the legal framework has expanded through the Information Technology (Intermediary Guidelines and Digital Media Ethics Code) Rules, 2021, the Digital Personal Data Protection Act, 2023, the Bharatiya Nyaya Sanhita, 2023, the Bharatiya Nagarik Suraksha Sanhita, 2023 and the Bharatiya Sakshya Adhiniyam, 2023. The latter three criminal-law enactments came into force from 1 July 2024, creating a new legal environment for substantive offences, criminal procedure and electronic evidence. The present research critically analyses the adequacy of India's existing cyber-crime framework in dealing with emerging forms of cyber offending. intermediary responsibility and institutional coordination. The paper argues that the principal challenge is no longer merely the absence of cyber-crime provisions, but the fragmentation, technological obsolescence, investigative capacity gap and coordination gap within the existing framework. The paper therefore proposes legislative harmonisation, specialised cyber courts, stronger digital-forensic infrastructure, standardised electronic-evidence protocols, enhanced international cooperation, AI-specific safeguards, victim-centric financial-fraud mechanisms and continuous legal reform.
Comparative Thermal Performance of Perforated Absorber Shapes in a Solar Air Heater
Authors: Research Scholar Nitin Kumbhare, Assistant Professor Dr. Pranay A. Bagde
Abstract: This paper compares semi-circular and parabolic perforated absorber shapes for a forced-convection solar air heater. Regular and zigzag patterns in inward and outward orientations were assessed from 11:00 to 15:00 using temperature, airflow and solar-irradiance data. A simple numerical comparison was also used to check the observed performance trend. The parabolic zigzag-outward absorber gave the best result, with 28.70% measured efficiency and a heat-transfer rate of 75 W/m² at 14:00. At the same condition, the semi-circular zigzag-outward design reached 27.07% efficiency and 53 W/m². The numerical estimate for the best parabolic case was 31.10%, which was 2.40 percentage points above the measured value. The results show that curved shape, staggered placement and outward orientation can improve air mixing and useful heat output without making the collector arrangement unnecessarily complex.
A Systematic Review Of IoT, Machine Learning, And Edge Computing For Real-Time Industrial Hazard Detection
Authors: Ashwini Vinod Waghale, Dr. K. P. Yadav, Dr. Salim A. Chavan
Abstract: There are several types of threats to human safety and industry that include gas leakage, excess temperature, vibration, fire, and machinery malfunction. This literature review aims at assessing the role of Internet of Things (IoT), Machine Learning (ML), and Edge Computing techniques in real-time industrial threat detection and monitoring. The literature review evaluates the use of multi-sensor system, machine learning for classification, Raspberry Pi based edge computing, MQTT communication protocol, and web-based monitoring systems. This paper focuses on Random Forest based classification of hazards, real-time monitoring of sensors, AES-256 Encryption-Decryption, authentication, secure MQTT communication, Flask dashboard, alert generation and relay based machine shutdown techniques. This paper discusses some of the challenges that include lack of real time validation, use of static data, sensor variations, secure data transfer, and combination of intelligent predictions with safety response mechanism. There is scope for future research in areas of secure IoT communication, edge based machine learning, real-time data processing, encryption and industrial safety response mechanism.
DOI: https://zenodo.org/records/22975804
Absence Of Evidence Is Not Evidence Of Susceptibility: Failure Modes Of Determinant-Count Rules In Resistome-Guided Antibiotic Selection
Authors: Seshu Kumari K, Dr. Munesh Kumar Sharma
Abstract: Frameworks that convert resistome content into antimicrobial guidance must decide what to do when no resistance determinant is found. The decision is easy to get wrong, and the error is invisible at the scale at which such frameworks are usually run. This paper audits, decision by decision, the 96 class-level outputs of a six-stage resistome-to-guidance framework applied to ten reference strains of the human gut microbiota carrying six determinants, against standard clinical microbiological references. Under a binary rule in which a determinant count of zero produces the decision preferred candidate, 89 of 96 decisions were positive recommendations. Eight classes of error were identified, several contradicting firmly established fact: cephalosporins were recommended for Enterococcus faecalis V583, which is intrinsically resistant to the entire class through low-affinity penicillin-binding proteins; aminoglycosides were recommended for both Bacteroides strains, which cannot accumulate the drug for want of oxygen-dependent uptake; and penicillins were recommended for both Enterobacterales despite a curated β-lactamase mapping to the penam ontology class, an implementation defect distinct from the conceptual one. Twelve of the sixteen clinical classes were unrepresented in the determinant panel and returned zero for every strain. A revised three-state rule, supplemented by a separately compiled intrinsic resistance register, removes every identified error; it converts 89 positive recommendations into 23 avoidance decisions on register evidence and 66 explicit statements of absence of evidence, and emits no positive recommendation at all. We argue that this last property is a requirement rather than a deficiency of any determinant-based guidance system, and that small auditable panels are a necessary stage in the validation of such systems.
DOI: https://zenodo.org/records/22977211
Security and Privacy Challenges in Virtual Classroom Platforms
Authors: Research Scholar Ms.Swati Mishra, Professor Dr.Kavita Mittal
Abstract: The scale of online virtual learning has grown rapidly, and various educational institutions are increasingly reliant on tools such as learning management systems, cloud services, web applications, and video conferencing platforms. These tools must process a large volume of sensitive information that cannot be disclosed casually, including education-related teaching data as well as the personal privacy information of teachers and students. This places extremely high security and privacy requirements on the entire virtual learning environment; if protection is insufficient, this information is highly vulnerable to leakage or tampering. Previous studies have identified authentication, access control, data integrity, content protection, encryption technology, and network security as core security concerns for online learning platforms [1]. Authentication confirms that the person logging into the system is the legitimate account holder; access control limits the scope of content that different users can view and operate on; data integrity ensures that information cannot be arbitrarily altered; content protection prevents course resources from being misappropriated and distributed; encryption technology and network security block vulnerabilities in data transmission and storage at the technical level. In recent years, systematic studies have specifically analyzed the cybersecurity issues faced by mainstream platforms including Moodle, Blackboard, and Microsoft Teams [2]. This framework integrates multi-factor authentication, Role-Based Access Control and Attribute-Based Access Control, secure communication, privacy-compliant data management, session monitoring, and security auditing functions. Its core goal is to build an extensible conceptual foundation that embeds security and privacy protection into every layer of the architecture, provide support for constructing reliable virtual classroom environments, and promote the responsible development of such environments.
Applications of Number Theory in Modern Cryptographic Systems: Foundations, Public-Key Schemes and Empirical Performance
Authors: Research Scholar Tripti Gautam, Professor Dr. Narendra Bahadur Singh
Abstract: Almost every public-key cryptosystem deployed today rests on a small set of results from elementary and algebraic number theory: modular arithmetic, Euler’s theorem, the Chinese remainder theorem, primality testing and the group structure of finite fields and elliptic curves. This paper explains how these results yield the RSA, Diffie–Hellman/ElGamal and elliptic-curve families, and then tests the central claims with reproducible experiments. Over 1,070 odd composites below 3,000, no Miller–Rabin base misclassified more than 23.7% of candidates (theoretical bound 25%), whereas the Fermat test was fooled by up to 76.6%. Baby-step giant-step discrete-logarithm cost followed the square-root law across 16–44-bit primes. Applying the Chinese remainder theorem made RSA private-key operations 2.8–3.8 times faster for 1,024–4,096-bit moduli. In a library benchmark, RSA-2048 signing was about 11 times slower than ECDSA P-256 signing, while RSA verification was several times faster than elliptic-curve verification. The paper closes with the parameter and implementation lessons that follow from these results and notes the limit imposed by quantum computing.
Sustainable Corporate Transformation: Leveraging Environmental, Social and Governance for Strategic Resilience and Long-Term Value Creation
Authors: Research Scholar Dr. Geo Paul K, Professor P. S. Aithal
Abstract: Corporate Governance is concerned with holding the balance among economic, environmental and social goals and between individual and communal goals. The corporate governance framework should be developed with a view to its impact on overall economic performance, market integrity and the incentives it creates for market participants and the promotion of transparent and well-functioning markets. The use of Environmental, Social and Governance (ESG) factors has shifted from being a niche topic in corporate responsibility to becoming a much more prominent role in strategic management, investment analysis and organizational accountability. This paper explores the potential of applying ESG principles to the company's strategy to achieve Sustainable Corporate Transformation and long-term company resilience. The research work uses a conceptual and secondary-research method that draws on academic literature, sustainability frameworks, regulatory trends, and environmental performance, social responsibility, governance effectiveness, stakeholder engagement, risk management and value creation. Corporate social responsibility is concerned with treating the stakeholders of the firm ethically or in a socially responsible manner. Corporate Transformation is a widest control mechanism within which a company takes it management decisions. Rebuilding of public trust and confidence by increased transparency in its financial as well as non-financial reporting and thereby increasing the shareholder value.
Ethical, Privacy and Fairness Safeguards for AI-Assisted Diagnosis of Depression and Anxiety: A Lifecycle Governance Framework
Authors: Mansi, Assistant Professor Dr. Grace M. BasanShrieh
Abstract: Artificial intelligence (AI) tools for detecting depression and anxiety process some of the most sensitive information a person can share: their thoughts, voice, movements and health history. Technical accuracy is therefore necessary but not sufficient for responsible use. This paper, derived from a doctoral study on AI-assisted diagnosis of depression and anxiety, proposes a lifecycle governance framework that translates four ethical principles (autonomy, beneficence and non-maleficence, justice, and explicability) into concrete safeguards at five stages: data collection, model development, validation, deployment and monitoring. Eight principal risks are identified and prioritised on a likelihood–impact matrix: privacy breach, consent failure, subgroup bias, automation bias, dataset shift, opaque output, miscalibration, and stigma or misuse. An analytic fairness demonstration shows how a sensitivity drop from 0.85 to 0.70 in a minority-language subgroup doubles missed cases from 22.5 to 45.0 per 1,000 people screened, even though the overall model appears acceptable. A privacy-preserving data pipeline and a checklist of minimum safeguards are proposed. The framework positions AI as accountable, clinician-supervised decision support and gives researchers, ethics committees and clinicians a practical basis for reviewing mental health AI projects.