Michael G. Email & Phone Number

At Stak, we have engineered a cutting-edge software solution designed to streamline the invoicing process for small business contractors. Our system adeptly handles invoice processing by automatically reading and categorizing documents according to custom cost codes. It ensures seamless integration with QuickBooks Desktop, facilitating the creation of accounts receivable and payable invoices, and constructing detailed budget-to-actual reports. Additionally, our software enhances operational efficiency by automating communications with vendors and clients, tailored to meet business needs. This comprehensive automation tool is crafted to optimize financial workflows and improve stakeholder engagement.

San Francisco, California, Us

St. Louis, Missouri, Us

St. Louis, Missouri, Us

Technical lead for corn fungicide response modeling efforts. Improve and enhance machine learning systems dependent on data from National Oceanic and Atmospheric Administration’s advanced global climate prediction model. Collaborate with operations on analytical reporting to aid enterprise trajectory.Boosted predicted fungicide response by 80% using a hierarchical Bayesian model coded in TensorFlow-probability -Unified multiple datasets into single point for faster analysis-Explored data to build a hierarchical Bayesian statistical model to predict field-level fungicide response -Evaluated against a linear mixed model and a gradient boosted tree, XGBoost, and presented results to key stakeholdersDeveloped a python package to consolidate and aggregate data to assist modeling efforts across teams-Engaged with researchers and stakeholders to identify the pain points this package would solve-Unified multiple database sources for use at a single point for faster analysis-Extensive review and testing of the codePresented dataset analyses to leadership to enhance corporate direction following company merger-Performed exploratory analysis of multiple datasets using Seaborn and matplotlib packages-Engaged with researchers from parent company to engineer data transfer and gain domain expertise-Assembled comprehensive presentation for leadership to encourage corporate focus in key areas

St. Louis, Missouri, Us

Increased weather forecast times from 18 to 120 hours by refining global temperature inversion prediction models-Created data pipeline for ingesting new weather data to accelerate response times-Cleaned and aligned weather data for model analysis-Trained, tuned, and validated model, enabling for temperature prediction inversions for entire planet

Pullman/Spokane/Tri-Cities/Vancouver/Everett/Global, Washington, Us

Orchestrated model development for pathogen-prediction system of Monilinia vaccinii-corymbosi fungus (Mummy Berry). Tested suite of machine learning models for agricultural industry. Refined and fine-tuned data infrastructure to ensure optimal application performance. Project published in academic journal Phytopathology in 2017.Produced machine learning model to predict fungal outbreaks resulting in an AUC score of 0.93 on a holdout year-Cleaned data from field researchers to weed out irrelevant information-Prototyped five different machine learning models: regularized logistic expression, multivariate adaptive regression splines, multilayer perceptron network, random forest, and XGBoost-Validated, retrained, and tuned results to ensure peak model performance for prediction of spore transmissions

Redmond, Washington, Us

Developed multidimensional time-series convolution neural networks to forecast agriculturally relevant weather variables. Invented sensor placement algorithm for agricultural use to reduce costs for consumers. Designed comprehensive software library to tune convolution neural network’s hyper-parameters automatically. Earned Top 40 placement out of 1900 applicants for Microsoft Research AI Residency Program.Reduced weather station costs by developing fast-processing neural network to predict forecasts-Scraped weather data from multiple sources to feed network for model training -Constructed base model architecture using one-dimensional convolutional neural network-Increased speed of model tuning tenfold by developing automatic hyper-parameter tuning algorithmLowered consumer costs by developing efficient sensor placement algorithms-Reduced client expenses by enabling swift data exchange between fewer sensors

Pullman/Spokane/Tri-Cities/Vancouver/Everett/Global, Washington, Us

Helped teach the soils analysis lab three times becoming proficient in all aspects of soil testing. Used the soil testing skills learned teaching this class to calibrate my own organic potting mix recipe using ingredients including blood, bone, kelp meal, alfalfa, and my homemade worm castings. This potting mix has produced some of the tastiest strawberries, tomatoes and potatoes I have ever eaten!

Us

I learned aseptic technique doing basic human tissue culture work in addition to basic high performance liquid chromatography analysis. I worked here while finishing my undergrad at UCSD and it was this job that led me to my Ph.D at Washington State

Master’S Degree, Data Science

Non-Matriculated, Mathematics

Doctor Of Philosophy (Ph.D.), Soil Chemistry

Bachelor'S Degree, Cognitive Science / Pre Med