Prem Kumar

Projects:

Multi-Modal Fraud Detection System:

• Developed a comprehensive fraud detection system integrating text and image data.
• Implemented NLP techniques (tokenization, stopword removal, TF-IDF) to preprocess textual descriptions.
• Employed image processing methods for loading, resizing, and preprocessing image proofs.
• Built a deep learning model combining CNNs for images and fully connected layers for text, achieving high accuracy in fraud detection.
• Tools: Python, TensorFlow, OpenCV, NLTK, Scikit-learn, Pandas, NumPy.

Impact:

• Improved the accuracy of fraud detection by leveraging multi-modal data sources, leading to more reliable identification of fraudulent activities.
• Enhanced the efficiency of the fraud detection process, reducing the time and effort required for manual review.

Logistics Cost Forecasting System:

• Developed a predictive analytics model to forecast logistics costs, including transportation, warehousing, and labor expenses.
• Collected, cleaned, and preprocessed historical data to ensure accuracy and completeness.
• Engineered relevant features (distance, shipment volume, fuel prices, seasonal trends) to enhance model predictive power.
• Designed and implemented various forecasting models (ARIMA, SARIMA, Random Forest, Gradient Boosting) for cost prediction.
• Evaluated models using metrics like Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and R-squared; selected the best-performing model.
• Integrated the forecasting model into the existing logistics management system for real-time cost predictions.
• Created interactive dashboards and reports using Tableau and Power BI for visualizing forecasted costs and trends.
• Tools: Python, Pandas,NumPy, Scikit-learn, Statsmodels, Tableau, Power BI.

Impact:

• Improved accuracy of cost forecasts by 30%, leading to better budget management and cost savings.
• Enabled proactive decision-making by providing timely insights into future logistics expenses.
• Enhanced resource allocation efficiency, reducing over- and under-utilization of logistics resources.

• Performed data manipulation and analysis: Leveraged Python libraries such as Pandas, NumPy, and SciPy to conduct complex data manipulation and statistical analysis, providing valuable insights and data-driven recommendations.
• Collaborated with cross-functional teams: Worked closely with various departments to understand their data requirements and provide customized reporting solutions that met their specific needs.
• Ensured timely and accurate delivery of reports: Managed report generation schedules and automated processes to guarantee the punctual delivery of accurate reports, enhancing decision-making capabilities across the organization.
• Developed and maintained MIS reports: Created comprehensive Management Information System (MIS) reports tailored to the needs of various stakeholders, ensuring they had access to critical insights and data.
• Developed automated data processing pipelines: Leveraged Python and Pandas to read and process large datasets, ensuring accurate and efficient data filtering based on specific vendor requirements.
• Optimized data workflows: Streamlined data processing workflows to improve efficiency and reduce processing time, resulting in faster report generation and delivery.
• Script optimization and maintenance: Regularly updated and optimized scripts to improve performance and adapt to changing requirements, ensuring seamless operations and minimizing errors.

• Assist in developing and implementing machine learning models
• Conduct exploratory data analysis with a focus on machine learning applications
• Work with tools like Pandas, NumPy, and scikit-learn
• Learn and apply statistical and machine learning techniques
• Collaborate with senior data scientists on more complex tasks.

• Dealing with Invoices like PO invoices and non-PO invoices
• Analyzing the Invoice with Two way & Three-way matching
• Discarding the invoices which are duplicates to the one which is already paid through same invoice
• Payment made to the vendor within deadline
• Exporting the report and forwarding the productivity report to the all-team members.