Bhavani Priya R

At TU Dresden, I developed Python pipelines for acquiring and processing DIGIT tactile sensor data to generate point clouds for 3D surface analysis. I also designed a calibration and data acquisition pipeline for the DIGIT360 sensor, integrating fisheye calibration, IMU alignment, and image processing for depth reconstruction.

As part of tactile perception research, I generated and analyzed more than 500 synthetic samples using the TACTO simulator and developed a diffusion-model-based data augmentation framework to generate realistic synthetic tactile data for training robust machine learning models. In my current research, I am developing a PyTorch-based pipeline for real-time surface-normal estimation and depth reconstruction.

Before my Master's, I worked as a Software Engineer at Teradata India Pvt. Ltd., where I contributed to migrating a monolithic application to Spring Boot microservices using Java and Hibernate, reducing deployment time by 50% and improving scalability by 40%. I also developed backend services and maintained enterprise software modules.

In addition, I fine-tuned BERT and RoBERTa transformer models for multi-label emotion classification, achieving an average F1 score of 0.79, and applied SHAP to interpret model predictions.

I enjoy solving challenging engineering problems by combining software engineering with machine learning research. I am particularly interested in developing reliable AI systems, computer vision applications, and data-driven software solutions, and I am currently seeking opportunities in Machine Learning, Computer Vision, Artificial Intelligence, and Software Engineering.