Case Study · Data Science · Statistical Modeling · Decision Systems

Rubin Variable Star Classification

Developed analytical workflows for large-scale astronomical time-series data through the NSF-DOE Vera C. Rubin Observatory research program. The project combined statistical exploration, feature engineering, and machine learning to identify and classify variable stars, transforming complex observational data into structured and interpretable research outputs.

Research impact

3.5 TB

Large-scale astronomical time-series data processed and analyzed through computational research workflows.

94%

Classification accuracy achieved by the initial variable star classification model.

3 Researchers

Supervised three undergraduate researchers, supporting their technical development and contributions to the research effort.

Confusion matrix showing the performance of the variable star classification model across astronomical object classes.
Confusion matrix for the initial variable star classification model. The evaluation provided insight into classification performance across object classes and helped establish a foundation for repeatable candidate identification.

The problem

Large astronomical surveys generate enormous volumes of observational data. Identifying variable stars within these datasets requires researchers to distinguish meaningful variability patterns from noise, incomplete observations, and other sources of uncertainty.

The challenge was not simply processing more data. It was developing a reliable analytical approach to identify relevant signals, extract useful features, and classify astronomical objects in a way that could support further scientific investigation.

Analytical approach

  • Processed and explored large-scale astronomical time-series datasets using Python-based analytical workflows.
  • Conducted statistical exploratory data analysis to understand observational patterns, distributions, and variability signals.
  • Applied feature engineering to extract meaningful characteristics from time-series observations.
  • Developed machine learning classification workflows to distinguish variable star categories.
  • Evaluated model performance using classification metrics and confusion matrix analysis.
  • Organized analytical outputs and documented workflows to support reproducibility and continued research.

Research collaboration

Beyond the technical work, I supervised three undergraduate researchers, helping coordinate analytical contributions and support their development within the research environment. This experience strengthened my ability to communicate technical concepts, collaborate across experience levels, and contribute to shared research objectives.

From scientific data to decision systems

Although the application was astrophysics, the analytical challenges are relevant across many industries. Large datasets are often noisy, difficult to interpret, and too extensive for manual review. Organizations face similar problems when identifying patterns, evaluating opportunities, and deciding where to direct limited resources.

This project reinforced an approach I bring to analytics more broadly: build structured workflows, make analytical methods interpretable, evaluate results rigorously, and turn complex information into outputs that people can actually use.

Research experience that scales beyond research

This experience helped develop my technical foundation in large-scale data analysis, statistical modeling, and collaborative problem-solving. Those capabilities continue to inform my work in business analytics, strategy and operations, and applied decision-support systems.