Applied data work — from multi-source ETL pipelines to model comparison and classification.
SourcesWrangleAnalyzeDashboard
SpaceX Launch Data Pipeline & Landing Prediction
Multi-source ETL and landing-outcome prediction for Falcon 9 launches
An idempotent data pipeline that collects, transforms, and analyzes Falcon 9 launch data from three independent sources, then feeds an analyst dashboard with dynamic visualizations and a landing-outcome classifier.
Consolidated launch records from the SpaceX REST API, Wikipedia scraping, and a SQL database into one standardized dataset of 180+ launches across 10+ validated fields
Balanced the training set with SMOTE before building an ETL pipeline (collection → wrangling → EDA/SQL → ML → dashboard), with Supabase as the storage layer and Plotly/Streamlit for the analyst-facing dashboard
Trained a landing-outcome classifier and surfaced launch-site geospatial patterns with Folium, translating results into trend summaries for reporting
Classification pipeline for next-day rainfall prediction
A preprocessing and classification pipeline built on mixed-type Australian weather data, evaluated for real-world interpretability rather than just model accuracy.
Built a preprocessing pipeline (ColumnTransformer, scaling, one-hot encoding) to clean and standardize mixed-type weather data
Tuned Random Forest and Logistic Regression models with GridSearchCV and StratifiedKFold cross-validation
Evaluated results with confusion matrices, classification metrics, and feature analysis, translating output into clear, plain-language performance insights instead of model-only reporting
Pythonpandasscikit-learn
Regression & Model Optimization Suite
Comparative regression modeling for high-dimensional data
An exploratory analysis and model-comparison workflow built to find the most reliable regression fit across high-dimensional and emissions datasets, with results packaged for non-technical review.
Analyzed linear, Ridge, and Lasso regression models on emissions and high-dimensional datasets using EDA and feature selection
Evaluated model fit with R², MAE, MSE, and RMSE, applying PCA-supported analysis to identify key variables
Built reusable preprocessing and tuning workflows to keep the analysis repeatable, then summarized findings for non-technical stakeholders