Detail-oriented MSc Data Science student with robust practical experience in machine learning, computer vision, and data analytics. Adept at developing AI-driven solutions—from designing real-time object detection system…
Detail-oriented MSc Data Science student with robust practical experience in machine learning, computer vision, and data analytics. Adept at developing AI-driven solutions—from designing real-time object detection systems using YOLO and PaddlePaddle for safety compliance to crafting predictive models for credit risk analytics and optimizing legacy code for manufacturing insights. Proven ability to translate complex data challenges into actionable insights through hands-on projects and diverse internships, leveraging advanced technical skills in Python, OpenCV, and synthetic data generation. Passionate about applying innovative, data-driven strategies to solve real-world problems and drive impactful business outcomes.
Designed and implemented AI-based safety compliance systems using computer vision, leveraging YOLO and PaddlePaddle for real-time object detection and activity recognition. Integrated detection models into workplace safety protocols to identify PPE compliance and human activity patterns. Developed real-time video processing solutions using OpenCV and Python to trigger violation alerts, enhancing workplace safety. Collaborated with cross-functional teams to deploy scalable AI-driven solutions. Gained hands-on experience in training deep learning models for real-world safety applications. Utilized advanced AI frameworks to improve compliance monitoring and automate safety assessments.
Designing and implementing AI-based safety compliance systems using object detection frameworks.
Collaborating with cross-functional teams to integrate real-time detection models into workplace safety protocols.
Utilized YOLO (You Only Look Once) and PaddlePaddle frameworks to train object detection models for PPE kits and human activity recognition.
Implemented OpenCV and Python for real-time video processing and violation alert mechanisms.
Gained hands-on expertise in deploying scalable computer vision solutions and PaddlePaddle framework for activity recognition.
As a Data Science Intern at PreludeSys India Pvt Ltd (May 2024 – Jun 2024), I developed and optimized predictive models to assess credit risk for both loan defaulters and credit card applicants. I leveraged the Synthetic Data Vault (SDV) to generate synthetic datasets, effectively addressing data imbalance issues and enhancing model generalization. Through detailed feature engineering, I identified key predictors of default behavior and trained multiple machine learning algorithms—including Logistic Regression, Random Forest, Gradient Boosting, and XGBoost—while fine-tuning hyperparameters using cross-validation techniques. This rigorous approach resulted in high-precision predictions, with F1-scores reaching 0.734 for Loan Eligibility, 0.836 for Loan Default, and 0.954 for Credit Card Default, thereby significantly enhancing decision-making for loan approvals. Tools and technologies such as Python and Scikit-learn were instrumental in ensuring the robustness and reliability of the models, contributing to more informed, data-driven financial risk assessments.
Developing predictive models to assess credit risk for loan defaulters and credit card applicants.
Addressing data imbalance and optimizing model performance through advanced synthetic data generation.
Achieved high-accuracy metrics in predicting defaulters, enhancing decision-making for loan approvals.
Leveraged SDV (Synthetic Data Vault) to generate synthetic datasets, mitigating class imbalance and improving model generalization.
Performed feature engineering to identify critical predictors of default behaviour, enhancing model interpretability.
Trained and evaluated multiple machine learning algorithms (e.g., Logistic Regression, Random Forest, Gradient Boosting) to select the optimal model.
Fine-tuned hyperparameters and validated results using cross-validation techniques to ensure reliability.
Achieved high-precision predictions across three critical risk models: Gradient Boosting (Loan Eligibility, F1: 0.734), XGBoost (Loan Default, F1: 0.836), and Random Forest (Credit Card Default, F1: 0.954), enabling data-driven lending decisions.
Enhanced model generalization by generating 10 lakh synthetic rows using SDV, addressing class imbalance and improving robustness on unseen data for reliable risk assessment.
As a Data Analytics Intern at Hyundai Motor India Limited (HMIL) (Jun 2023 – Jul 2023), I modernized legacy R scripts by migrating them to Python, significantly enhancing code maintainability and accessibility for engineering teams. I collaborated with cross-functional teams to seamlessly integrate the revamped code into production workflows and designed an interactive QlikSense dashboard for real-time monitoring of Work-in-Progress (WIP) buffer counts. This dashboard streamlined production planning by providing dynamic visualizations and predictive alerts for buffer time thresholds. Through systematic analysis and optimization using pandas and NumPy, I reduced computational overhead and memory dependency, which improved runtime efficiency for high-volume manufacturing data processing. This project not only ensured 100% functional parity between the original R logic and the new Python implementations but also enabled faster debugging and scalable solutions for ongoing process optimization initiatives.
Migrating legacy R programming scripts to Python to improve code maintainability and accessibility for engineering teams.
Collaborating with cross-functional teams to ensure seamless integration of converted code into production workflows.
Designing and deploying an interactive QlikSense dashboard to monitor real-time Work-in-Progress (WIP) buffer counts and streamline production planning.
Successfully optimized Python scripts to reduce memory dependency, enhancing runtime efficiency for high-volume manufacturing data processing.
Conducted systematic analysis of R scripts to replicate logic in Python, ensuring 100% functional parity while improving code readability.
Leveraged pandas and NumPy for data processing optimizations, reducing computational overhead in Python implementations.
Developed an interactive QlikSense dashboard with dynamic visualizations to track WIP buffer counts, integrating predictive alerts for buffer time thresholds.
Delivered Python-equivalent scripts adopted by engineering teams, enabling faster debugging and future scalability.
Launched a real-time WIP buffer monitoring system, reducing manual reporting efforts and improving response times to production bottlenecks.
نظام التوظيف التشغيلي المعتمد على الذكاء الاصطناعي أولاً، الذي يوحّد ويؤتمت دورة التوظيف بالكامل للمؤسسات الحديثة حول العالم.
د.جوب منطقة حرة ذ.م.م. 2026 © جميع الحقوق محفوظة
