Nishita Sherathiya
Columbus,OH
Summary
Accomplished Data Science Researcher at the University of Dayton, specializing in machine learning and computer vision. Developed innovative predictive models and real-time dashboards, enhancing data-driven decision-making. Proven ability to improve risk identification accuracy and streamline data processes, demonstrating strong analytical skills and a commitment to excellence.
Overview
5
5
years of professional experience
Work History
Data Science Researcher
University of Dayton
2024.09 - Current
- Developed an NSF-funded, real-time hazard detection system using a Transformer based self-calibration method to convert 2D site images into 3D coordinates, achieving a 3.471cm Mean Absolute Error (MAE) in tool elevation calculations to proactively mitigate falling-object risks from heights in construction sites.
- Engineered a hybrid computer vision and statistical framework to evaluate edge-based tool stability achieving a 2.5% MAE, to proactively flag falling-object risks across active construction sites.
- Designed lightweight texture classification system using augmented LBP descriptors with local entropy and standard deviation features, achieving 78% classification accuracy on DTD-47 texture dataset at 56x lower dimensionality than ResNet-50 features, enabling real-time inference with SVM classifier validated via 5-fold cross-validation.
- Developed coma recovery outcomes from 3D volumetric MRI scans using DenseNet121-3D in PyTorch and MONAI, achieving 84% AUC on 20-patient clinical dataset with leave-one-out cross-validation and GradCAM overlays to enhance interpretability of results.
- Extended Gaussian Nonlinear Line Attractor (GNLA) network to evaluate robustness under Gaussian noise, occlusion, and missing features on MNIST dataset, benchmarking against NLA, PCA, and Autoencoder baselines, achieving 0.91 attractor stability score, 78% classification accuracy under occlusion, and 0.81 SSIM reconstruction, outperforming all baselines on degraded input conditions.
Data Analyst
MRO (India) Pvt. Ltd
2021.07 - 2024.05
- Analyzed 250K+ patient records to identify high-risk cases, building and refining a predictive scoring model that improved risk identification accuracy by 30%, used statistical techniques to address data imbalance, actionable results for care teams.
- Designed and deployed a real-time Power BI dashboard tracking request volume, turnaround time, and revenue by hospital account, integrated with SQL pipelines, enabling finance team to prioritize outreach.
- Developed automated data validation framework using Python with schema checks, deduplication, and business-rule validation to identify and resolve data discrepancies, enhancing data quality and achieving 80% stakeholder satisfaction.
- Built an automated data pipeline in Python to structure and extract insights from 150K+ unstructured records, reducing manual review time by 50% and automating audit workflows to support compliance reporting.
- Designed and deployed real-time Power BI dashboards integrated with SQL pipelines to track key operating metrics, enabling stakeholders to monitor performance and prioritize action across 5,000+ records with reduced manual review time.
- Executed reconciliation and validation of 250K+ records in SQL for regulatory compliance (HEDIS, HIPAA), decreasing data quality errors by ~15% and facilitating accurate audit-ready reporting.
Education
MS - Computer Engineering
University of Dayton
2026-05
BE - Computer Engineering
Gujarat Technological University
2020-05
Skills
- Computer vision
- Machine learning
- Predictive modeling
- Statistical modeling
- Feature engineering
- Model validation
- Data analysis
- Python programming
- SQL querying
- Power BI dashboards
- Big data analytics
- Time series analysis
- Optimization techniques
- Natural language processing
Research Paper
- Monocular vision-based localization and risk assessment of falling object hazards on construction sites (In progress)
- Real-Time Augmented Reality Framework for Construction Safety Training Using Object Detection and Spatial Mapping (submitted)
Timeline
Data Science Researcher
University of Dayton
2024.09 - Current
Data Analyst
MRO (India) Pvt. Ltd
2021.07 - 2024.05
MS - Computer Engineering
University of Dayton
BE - Computer Engineering
Gujarat Technological University