Shruti Balaji

Implemented and fine-tuned large-scale audio TTS models using LoRA and PEFT to add emotional layers and evaluated performance across diverse accents; optimized training and inference with CUDA-accelerated PyTorch on multi-GPU setups.

Developed Faster R-CNN and multimodal fusion pipelines to detect tumors and segment breast cancer on DICOM mammograms, extracting ROIs and improving diagnostic localization across modalities.

Performed data cleaning, visualization and time-series forecasting (ARIMA) on offshore wind employment data, predicting a 7% jobs increase by 2025 and delivering stakeholder reports and Tableau BI dashboards.

Built backend components in C# and C++, implemented multithreading simulations, and developed React front-end features with MySQL integration while using Microsoft Azure for cloud development and deployment workflows.

Developed CNN, ResNet50 and Faster R-CNN models for breast cancer classification and segmentation achieving 94.6% accuracy on CNNs and reduced training time via parallelized supercomputer training; built Random Forest pipelines for inhibitor screening with 87% accuracy.

Designed interactive Tableau dashboards from call-center data to surface KPIs (call volumes, AHT, department demand) and used R for statistical modeling to enable real-time monitoring and operational improvements.

Activities and societies: Researcher at Multi-scale Medical Robotics Lab; projects in breast cancer segmentation, multimodal fusion, and large-model pipelines.

Pursuing a Master of Science in Data Science with coursework and research in machine learning, computer vision, and large-scale model fine-tuning.

Activities and societies: Machine Learning intern at Computational Lab; developed CNN/ResNet/Faster R-CNN models and Random Forest screening pipelines.

Completed a Bachelor of Technology in Bioinformatics with projects in deep learning for cancer classification, segmentation, and compound screening.