Work and Research Projects
I wrote a convolutional neural network to practice and learn PyTorch (with GPU acceleration) that trained on the CelebA dataset. The purpose is to recognize facial attributes. The network was produced from this paper.
While I worked at Arculus from May – July 2017, I had the opportunity to develop a logging system for the autonomous guided vehicles (AGVs). I developed an architecture for the AGV to communicate wirelessly to a server where logs could be dumped in close to real-time for a monitoring service. I wrote Python code to wirelessly serialize data and transmit it to an ELK server using ZeroMQ. Check out the Arculus website and all of the cool things they are doing!
At the Technische Hochschule Ingolstadt, I worked on developing a battery testing lab with PhD candidate Fabian Steger. For this work I was published at AAEE 2017.
From August to December 2016 I worked on using Raspberry Pi nodes as robot controllers to distribute a wireless OLSR mesh network throughout a building or across a general area. I worked with PhD student Paolo Regis on this project and won Goldwater Scholarship Honorable Mention for my proposal and research. As I traveled to Germany for the spring semester in 2017, I was unable to develop an autonomous system for deploying the nodes.
Extending the IP blacklist project, I continued work on developing a search to check which IPs had touched the network by cross-linking various system logs based on a list of IP addresses. A user could provide a list of IP addresses and use a web application to communicate with Elasticsearch to return logs and information about the list of IP addresses. I utilized a Kafka queue for the web requests, SiLK to analyze security data, and Google MDL for the front-end web page. I presented at the Annual Nexus Conference in 2016 based on this work.
On my first major research project, I helped to develop an automated system to gather a list of blacklisted IPs for the CISO of the University of Nevada. I worked with Daniel Lopez to parallelize a retrieval system that dumped the list into an ELK stack which was easily searchable in close to real-time (<1min). Out of this project I made a few videos for the STEMBOPS group and wrote a white paper on the ELK stack (in 2015).