Michael Hays

Banduro is a chatbot management platform that allows non-technical users to create their own chatbots and deploys them on their business webpage.

I designed an "injection" process, which allows a chatbot to be dynamically loaded into an iframe in the user's website.

To make this work, the user needs to copy and paste two

urPlus is a collection of tools to enhance the Udacity reviewer and mentor experience, including integrated remark storage, and retrieval.

I initially created the tool for my own personal use, but later released it publicly under the GPL-3.0 when I realized how useful it could be to the Udacity review community.

Within one month, the Chrome extension had over 120 downloads, which accounted for nearly 1/3 of all reviewers. In 2017, I moved on from the reviewing community, and transferred the project to another reviewer. I am still incredibly proud of how much it continues to help my fellow reviewers.

In 2017, I was involved in a contested case over the business requirements of Smart Meter Texas (SMT), which is a website used to store and retrieve energy usage data for Texas consumers.

Awesome Power Texas, which is a small business I began building in 2016, uses the data from SMT to determine the ideal electricity plan for Texas consumers through a predictive usage model.

End users and third parties were having trouble navigating the SMT web portal as well as initiating data access agreements, so the Public Utility Commission of Texas launched a contested case to improve the platform.

I participated by submitting a proposal on how to redesign pieces of the application at a high-level, in ways that would ease the registration and third-party integration process. In particular, I designed a better authentication system, which involved a secret key delivered to each energy consumer, and provided the security mechanism, which was absent in the original website.

Power to Choose is the state-run website for choosing an electricity plan in Texas. One major shortcoming of the website is that it only shows the expected cost for each plan at three cutoff points: 500 kWh, 1,000 kWh, and 2,000 kWh.

This means that the vast majority of the plan terms are unknowable without looking through every Electricity Facts Label (EFL) directly, of which there are hundreds. My website, Awesome Power Texas, collects the EFLs from Power to Choose, analyses the PDF text (either directly or through OCR), and creates a model of the entire range of the plan.

This is necessary to calculate the expected plan cost for a particular user, using their actual historical usage data. To store the information ascertained from the EFLs, I mapped each plan as a piece-wise function of charges (flat fees), and rates (applied per kWh), and stored this information as a nested array in our Postgres database.

The attached code demonstrates how we use that information to calculate the plan's cost at any usage level, and not just the cutoffs provided by Power to Choose.