DustyCult is a Julia code for modeling transient stellar dimming from circumstellar dust occultation. It includes stellar parameters, limb darkening, stellar grids, dust shapes, opacity laws, sky-plane motion, and forward light curve models.
The package supports inference with Turing, JSON-configured light curve fitting, and Parquet outputs for posterior samples, light curves, predictive intervals, predictive fluxes, and run manifests.
I participated in the Linking Writing Processes to Writing Quality Kaggle competition, concluding in January 2024. I engineered dozens of features from provided keystroke log data based on recent computational linguistics and psychology literature, applied scikit-learn models, analyzed feature importance and model performance, and placed in the 63.3% on the Efficiency Leaderboard.
This poster summarizes my solution.
Prototyped an updated cleanroom garment with tear sensors, improved boot covers, and redesigned masks for use in Intel's semiconductor fabrication plants; designed with Arduino boards and cleanroom garment materials; won 1st place in the competition.
Researched performance of various solar car geometries, designed canopy and aeroshell in SolidWorks, ran CFD with STAR-CCM+. Meshed existing canopy with photogrammetry software, compared prototypes to physical model, finalized design, integrated with remainder of solar car.
Collaborated with 10-person interdisciplinary team to conceptualize a mission to drill water-ice from the lunar south pole, including site selection, rover design, system integration, risk mitigation, and success criteria. Modeled entry, descent, and landing of a lunar rover; prototyped compact lunar regolith drill; formulated a CONOPS; budgeted instrumentation; produced a Preliminary Design Review.