Jon Zink Email & Phone Number

Mountain View, Ca, Us

Mountain View, California, United States

Pasadena, California, United States

Implemented K-means clustering analysis with Python/Scikit-learn to group planet classes, reducing measurement uncertainty by 22%. Deployed Bayesian hierarchical modeling with PyMC to resolve a 30-year-old planetary formation mystery by identifying intra-system correlations in giant planet systems. Developed Gaussian process regression methodology for robust correlated-noise removal, improving planetary signal recovery by 9X with the method adopted by 12 independent research groups. Executed conditional survival analysis in R to validate 60 planet signals, concluding each event had a false-positive likelihood less than 0.1%.

Los Angeles, Ca

• Designed Python algorithm able to search highly contaminated time series data and identify planet signals with 94% reliability, resulting in the discovery of 372 planets: see Forbes & Newsweek press.• Implemented random forest regression with TensorFlow to characterize ∼ 200, 000 stars, resulting in 93% accuracy of the validation set classification, a 48% improvement of existing methods.• Derived Poisson point-process expansion, addressing bias issues associated with order statistics, improving the precision of inferred exoplanet population models by 36%.• Developed novel likelihood function for forward model comparison, using the Anderson-Darling EDF and a modified Poisson PDF, enabling algorithm convergence 10X faster than existing methods.• Produced open source Python packages, used by 27 independent research groups: see EDIunplugged & ExoMult on Github.

Doctor Of Philosophy - Phd, Astrophysics

Bachelors, Astrophysics