Ted Bohn Email & Phone Number

Greater Seattle Area

• Implemented automated classification (unsupervised and partially supervised) of device usage episodes and evaluation of equipment condition from power consumption time series measured at smart sockets, using Python on AWS, Docker, Kubernetes, EKS. Algorithms include KMeans clustering, Gaussian Mixture Model, Bayes Theorem, probabilistic ensemble forecasting.

Greater Seattle Area

Live performance of Brazilian music on trombone and percussion. Arrangement of horn parts.

Seattle

• Built CutoffPredictor, a tool for water utilities to predict customer defaults from payment and water usage behavior, as a consulting project for Valor Water Analytics.• Implemented complete pipeline in Python: querying 2M-record Amazon Redshift database via Postgre SQL; data cleaning and feature engineering in Pandas; oversampling with SMOTE; training and cross-validation of random forests and logistic regression algorithms from SciKit-Learn, Statsmodels, and H2O to recognize behaviors indicative of cutoff; and plotly/dash to display results in a dashboard.• Predicted cutoffs with ROC AUC of 0.92, and provided estimates of expected revenues lost on the order of $300k/y for a given utility.

Tempe, Arizona

Investigated impacts of changes in climate and land use on hydrology in the US and Mexico.• Developed PITRI, a novel Python algorithm for estimating hourly precipitation from daily totals, with parameters derived from linear regression against long-term climate data, leading to biases of only 1-2% in hydrologic model outputs over North America, compared to the previous widely-used algorithm’s biases of up to 30%.• Generated MOD-LSP, multi-decadal land cover parameters for hydrologic modeling over North America via Python (xarray) and ArcGIS/ArcPY, leading to 30% improvement in accuracy over arid regions and enabling studies of land use change impacts.• Created ensemble forecasting system over the Colorado River Basin using the improved precipitation algorithm and land cover data. This system was adopted by the Central Arizona Project and Los Alamos National Laboratory.

See description for Research Scientist at Arizona State University.

Investigated impacts of climate change on the hydrologic and carbon cycles.• Extended the UW surface water (drought) monitoring system to be a 5-model ensemble covering the US and Mexico, using C, Fortran, and Perl, reducing the impact of model bias. Inspired the NOAA National Drought Monitor to set up a similar framework.• Forecasted stream flow in major river basins across North America via hydrologic models, Bayesian parameter estimation, Bayesian model averaging, PCA, linear and logistic regression, and time series/harmonic analysis.

See description under Graduate Student Researcher at University of Washington.

• Built automated pipeline in Perl and MySQL to input, clean and analyze 1M+ observations of salmon migration on the Columbia and Snake rivers.• Automated prior manual data entry/management, saving $100K/y and eliminating the need for a full-time position.

Phd, Civil Engineering/Hydrology And Water Resources

Activities and Societies: Web Developer, American Water Resources Association, 2004-2006 UW Toastmasters (Competent Communicator).

Masters, Geophysics

Activities and Societies: UC Jazz Ensembles Afro-Cuban Ensemble, 1992-1993Tocher Memorial Fellowship, 1991-1992 Thesis Title: "Time-Lapse.

B.S., Engineering Physics

Activities and Societies: UC Berkeley Jazz Ensembles, 1987-1989UC Regents' Scholarship, 1987-1991