Komal B Email & Phone Number

Tempe, Arizona, United States

● Utilized python libraries including Selenium for web scraping and applied Natural Language Processing techniques to analyze and compare content, providing insights into trends, themes and discrepancies between Apple News+ outlets and local newspapers.

Tempe, Arizona, United States

● Assisted in teaching CIS 508: Machine Learning in Business and CIS 515: AI and Data Analytics Strategy.● Supported students in understanding machine learning algorithms, data analytics techniques, and strategic AI applications.

Hyderabad, Telangana, India

● Developed Spoof detection models leveraging convolutional neural networks in eKYC process used by the JiomHere app to provide new sim cards to customers by verifying whether the ID (Aadhar, Pan, Voter, Passport)provided is fake or real with 97.81% accuracy.● Enhanced the accuracy of Spoof detection Models by refining architectural elements within EfficientNetB0,MobileNetV1 and MobileNetV2 with TensorFlow and keras, leading to a 3-4% improvement.● Generalized and improved accuracy of spoof models by constantly identifying model behavior and by analyzingchanges in production data, achieving a 3%-5% accuracy increase in various spoof detection models.● Assisted in developing Emotion Detection models using EfficientNetB0 to analyze facial expressions of eachemployee image captured through the Jio mHere Pro app with 94.32% accuracy.● Contributed to developing and deploying in 11 Models including binary and multi-class classifiers utilizingsigmoid and ReLu activation functions with SGD and Adam optimization algorithms.● Implemented various data management and pre-processing techniques and Multi-Task Cascaded ConvolutionalNeural Networks (MTCNN) for better face detection, cropping and alignment.● Implemented class weighting and Bayesian Optimization which improved the accuracy of production models by2%-3%.

Hyderabad, Telangana, India

● Assisted in developing a Drowsiness Detection model employing MobileNet-v2 and ResNet-50 achieving 98.12%accuracy.● Built Tableau dashboard to visualize core business KPIs (e.g. Monthly Recurring Revenue), saving 10 hours per week of manual reporting work.