Human Robotic VIP

Predicting Gait Phase

---

The overall goal of our vertically integrated program was to design and develop lower limb powered prostheses. My subteam researched where to place sensors and which sensors would be best to use. We took that data and developed a machine learning algorithm to control the limb. By creating an accurate model, we hoped to reduce the weight of prostheses for those with physical limitations.

Close

Human Robotic VIP

Considering Real World Limitations

---

My specific task within my subgroup was to find a machine learning algorithm that had the ability to run real-time operations on an Arduino embedded on a prosthesis. We had to keep in mind resource consumption as well as heat generation because the prosthesis will be touching skin.

Close

Interned at

The Federal Reserve

Using Neural Nets to Categorize Payment Data

---

At the Federal Reserve Bank of St. Louis I was tasked with determining if an entity was a person or a business. I decided to use machine learning algorithms in SAS Enterprise Miner. Using a table with no categorization, I labeled 3,000 records and used it to train the models. I initially compared Trees and Neural Nets but it soon became apparent that, because of the nominal nature of most of the inputs, Neural Nets would be the better option. I tested single-layer and multi-layer both before and after running a principal component analysis. I determined that single-layer performed the best based on the average squared error, and, after grading 1,500 of its outputs, it achieved a success percentage of 92.47%. This project will allow for better analysis of payments and better detection of anomalous payments.

Close

VG Dev

Enemy Creation for the Game Memorai

---

This enemy was a mechanic that I contributed to the team. I designed and programmed how he would behave as well as the original art for him. This entity forced players to dodge by jumping, a tactic that had not been utilized in previous enemies. It forced players to think outside the box and created an overall better game.

Close

VG Dev

Spell Creation for the Game Bard-Mage Arena

---

This was the first team based project that I worked on. I contributed to the team by designing the pictured spell as well as 3D-modeling it. The finished game can be found here. With the addition of this spell the game developed a multitude of play styles, allowing for more interesting gameplay.

Close

Interned at

The Federal Reserve

Enhancing Automated Processes

---

In SAS Enterprise Guide, I improved the run time of an automated process to 80% of the original. I also improved the quality of the graphs produced by the automated process, allowing for the information to be read faster and more easily. This allows customers to quickly realize the impact our analysis has. Additionally, I added suggestions on how people may be defrauding the government as well as a base model for the analysis. I coded a test sample to demonstrate how their address data could be used to help detect anomalous streets that had over forty payments going to it. This allows them to identify even more payments that could be a source of fraud.

Close