Feb 2024 - May 2024

Data Scientist

• Researched Neural Network forgetting in distributed computing like FedAvg and Round Robin on diverse data. • Conducted performance study on 10+ Deep Learning models using TensorFlow, resulting in a 45% cost reduction . • Recognized as the ’Best Data Science Team’ at CKIDs USC for exceptional interdisciplinary research efforts.

Dec 2021 - May 2022

Computer Vision Engineer at Dimensionless Technologies Pvt Ltd

• Cross-collaborated to train EfficientNet-B5 Deep Learning model using Agile on Docker distinguishing counterfeit electronics, achieving 97% accuracy. • Trained YOLOv4 Object Detection models for 15 solar panel defects on Azure, achieving 55% mAP. • Utlilized OpenCV for image processing and built a Machine Learning model to optimize defect precision by 24%. • Conducted Grad-CAM analysis on CNN layers for model interpretability, boosting accuracy by 1.2%. • Peformed synthetic data generation using GANs for solar defect detection, decreasing false positives by 8%. • Developed ETL pipeline, transforming JSON to CSV from Google Drive to Azure, boosting efficiency by 3x.

Oct 2020 - Jan 2021

Data Scientist / Machine Learning Engineer at Skinzy Software Solutions Pvt Ltd

• Built computer vision Mask RCNN model in TensorFlow to highlight skin abnormalities with an IOU of 0.6. • Implemented Deep Learning ResNet-50 model to detect skin abnormalities, yielding an accuracy of 85%. • Analyzed collected data using Python and Excel for skin disease treatment recommendation system.

Oct 2020 - Dec 2020

Data Scientist Intern at Sardar Patel Institute of Technology

• Conducted a study on deep learning CNNs such as VGG16, ResNet50, and InceptionV3 using Transfer Learning for chest X-ray pneumonia prediction, achieving 98% recall and 94% accuracy. • Performed Image processing and Data Augmentation, to increase the data size by 5x. • Published research in the IEEE International Conference - DOI: 10.1109/I2CT51068.2021.9417872 .

Feb 2021 - Apr 2021

Machine Learning Engineer Intern at Sardar Patel Institute of Technology

• Led a team to program a Stacked Ensemble ML Model for EEG emotion detection by combining 8 ML models: neural networks, Random Forest, SVM, Logistic Regression, KNN, XGBoost, and LightGBM, achieving 97% accuracy. • Utilized Principal Component Analysis to reduce dimensionality by 94%, thus optimizing resources. • Published research in the IEEE International Conference - DOI: 10.1109/ICCCNT51525.2021.9579818 .