Podele Bhavani

Experience

• 

  GlobalLogic

  Mar 2020 - May 2022

  Geographic Information Systems Analyst

  • Efficiently managed over 50 GIS map services using ArcGIS Server, catering to both internal and external stakeholders, and performed intricate spatial data analysis for urban planning with ArcGIS Desktop. Enhanced operational efficiency by automating geoprocessing tasks using ArcGIS modelers, achieving a 30% reduction in processing time.• Development of raster and vector map layers in QGIS:• Crafted over 40 vector and 20 raster layers for key projects, including regional land use and environmental conservation, by meticulously adhering to municipal GIS standards and procedures, and ensuring precise data integration into GIS tables.• Inputted demographic data into GIS tables, creating layers that were used in urban development studies. Enhanced transportation network GIS layers with over 1000 miles of road data. Collaborated with urban planners to refine GIS workflows, resulting in a 25% system performance boost, and developed user-centric information and analytic solutions.• Developed and maintained data components for a real estate market analysis tool, providing insights into property valuation trends.• Managed and updated a comprehensive city-wide GIS database, maintaining a high accuracy rate of 98%, and ensured up-to-date utility network data, overseeing over 500 miles of water and sewage infrastructure. Developed interfaces for a natural resource management system to facilitate efficient data access for environmental analysis. Show less

• 

  Internshala

  Jun 2022 - Nov 2022

  Machine Learning Intern

  • Developed a Titanic Survival Prediction model, recognized in the top 10% on Kaggle.• Refined model accuracy to 85% by utilizing ensemble techniques like Random Forest, XGBoost, and SVM.

• 

  Saint Peter's University

  Aug 2023 - now

  • In the realm of data science, I’ve designed and refined models to optimize analysis and predictions, leveraging a blend of evolutionary algorithms and neural networks. My work resulted in a 20% boost in tuning efficiency and a 15% increase in predictive accuracy. Notably, a hybrid model I engineered improved parameter tuning for complex datasets, as evidenced by training and validation losses decreasing steadily over 90 epochs.• I’ve honed an adaptive parameter adjustment mechanism using Differential Evolution strategies, enhancing convergence and accuracy in large neural network models—a 25% reduction in convergence time and 30% in error rates attest to this. My strategic innovations also include a power function-based fitness evaluation within neural networks, driving a 40% surge in computational efficiency.• My foray into NLP with a Transformer-based model for classifying COVID-19 related tweets achieved a remarkable 94% accuracy and an F1 score of 0.92. I’ve improved data preprocessing, bolstered classification accuracy by 20%, and effectively managed class imbalances, securing balanced accuracy and model sensitivity. These efforts, coupled with deploying leading models like BERT, GPT, and XLNet, yielded precision up to 95% in multi-class categorization.• My dedication to public health discourse analysis through advanced NLP techniques has improved categorization efficiency by 30%, demonstrating societal impact. My collaborative spirit has propelled research group productivity forward by 25%, contributing to both academic circles and real-world applications. Show less • Pioneered an advanced algorithmic approach for large-scale spatiotemporal data analysis to identify and predict urban mobility threats using New York City taxi trip datasets.• Engineered machine learning models to process and analyze complex data, employing convolutional neural networks (CNNs) for spatial patterns and recurrent neural networks (RNNs) for temporal dynamics, achieving an R-• squared value of 0.6838.• Implemented anomaly detection techniques that flagged outliers in trip durations and geospatial hotspots, enhancing cyber-security measures for urban infrastructure.• Achieved model precision with a mean squared error (MSE) of 0.3989 and mean absolute error (MAE) of 0.4184, underscoring high predictive accuracy.• Employed TensorFlow and NetworkX for modeling and graph-based data visualization to elucidate intricate urban dynamics and potential security threats.• Formulated a graph network of interconnected events to reveal complex patterns and relationships in urban mobility, paving the way for optimized city planning and intelligent transportation systems. Show less

  • Graduate Research Assistant

    Feb 2024 - now

  • Graduate Research Assistant

    Nov 2023 - Feb 2024

  • Research Assistant

    Aug 2023 - Nov 2023