-Helped create the TestBank system, now used by NetMath to automatically generate relevant, unique exams and problem sets, and learns it's own problem selection algorithm based on user feedback.-Largely developed and oversaw the deployment of CAS-ILE, NetMath's interactive code-based math learning system, written in JS, C#, and Mathematica.-Automated several administrative tasks (courseware enrollment, status change, email updates, etc.) through the CAS-ILE api-Developing new client-side JavaScript framework to smoothly render highly interactive, visually compelling math demonstrations in the Mathable courseware app.

-Trained and led the development team for step-by-step math, and continued to improve its advanced math capabilities.-In 2016, step-by-step-math generated 90% of Wolfram Alpha Pro subscriptions.-Contributed to broad functional areas of Wolfram Alpha, natural language parsing, machine learning, vector calculus, and interactive visualization.Developed several of the functions in Mathematica's Complex Dynamics Package in Mathematica and C.

-Helped create Wolfram Alpha's popular step-by-step-math engine, along with many of its more advanced functions, including those for limit solving, discontinuity and critical point detection, and proving convergence/divergence of sums and integrals.-Created novel mathematical techniques to augment alpha's abilities, including a more general function decomposer, a detailed asymptotic germ classifier, a dominating function generator, and an algorithm to optimize the step-by-step solution of difficult limit problems.

-Taught both standalone classes and recitation sections of Calculus I-III.-Created interactive visualizations, exercises, and other content for UIUC's Calculus & Mathematica sequence as well as for my own classes.

-Designed and implemented neural networks, statistical classifiers, and other learning algorithms to enable ChemSensing's “smell-seeing” device to interpret the highly volatile and multidimensional dynamics of various arrays of aroma-sensitive colorimetric dyes.-Successfully trained the device to rapidly detect, identify, and quantify various types of aromatic compounds, such as toxic industrial chemicals or the metabolites of various bacteria in septic blood, and to dynamically evaluate its initial judgments as it collected more data.-Used unsupervised dimensionality reduction techniques and outlier detection to avoid overfitting the algorithm, even to noisy and/or highly underdetermined sets of training data.-Led the team of engineers which embedded purpose-trained classifiers into microcontrollers for specific medical and industrial deployments.-Conducted and later supervised the analysis of experiments, and prepared visualizations of the device's "olfaction" and identification of analytes for scientists, investors, and journal articles.