Five Rings Capital is a proprietary trading firm that merges strategy and technology to compete in global markets. I developed back-end mathematical engines to operate a high-frequency trading robot on the NYSE stock exchange, and optimized for maximum run-time efficiency while programmatically predicting the behavior of other trading robots and human traders. I also developed front-end UI web applications to allow for efficient and intuitive fast-paced trading. Since development at trading firms is private and copyrighted to the company, there is no code I can release.
Microsoft’s AzureML platform is a vastly powerful cloud-computing engine. It allows for distributed training of machine learning models, while melding with the capabilities of popular ML toolkits, such as TensorFlow and PyTorch. As the scale of ML advances, so does the necessity to construct pipelines of models, their data processing utilities, and their deployment requirements. I built the backend capabilities within the AzureML codebase at Microsoft to successfully deploy the first ever instance of KubeFlow on Azure. My Azure software was integrated into production. I also contributed installation patches to KubeFlow’s open-source codebase.
Jobcase is a job-searching and networking platform with over 100 million users. Unlike other popular websites, such as LinkedIn, Jobcase is aimed at individuals who may not have graduated high school or college, and are entering the workforce with little skills. By fostering conversation amongst users, we are able to match job-seekers with employed users knowledgeable about their company’s hiring techniques. I ideated, developed, tested, and put into production an MLFlow machine learning model to automate the moderation of these conversation threads, to liberate part of our sales team and allow them to focus on other tasks. I also developed and deployed a pipeline to optimally construct small groups of users into conversation “pods” that would help their situations. I also contributed to Apache Beam’s open-source Python SDK during development. Since Jobcase is not open source, I have no releasable code.
As a consultant for Winquest Cybersecurity Services, I developed a network protection software using BRO IDS and databases of known cyber threats. The software, intended for use on an isolated-network government entity, kept watch of all network activity. Using Python and BRO, I analyzed this intelligence data and corresponding metadata, and determined if any activity could be malicious. I then used these triggers to set up new protections against the identified known vulnerabilities and threats.
NASA Goddard develops technology for several robotics-oriented missions, such as ARRM and RESTORE-L. These require the robust testing of large-scale, powerful robotic arms, which necessitate the operator to process several monitors of sensor feedback in tandem with moving the arm. To increase the safety and efficiency of these tests, I developed augmented reality software for the Google Glass, that pulled sensor information packets from the testing network, processed the data, and presented visually intuitive graphical displays overlayed on the real-life robot. I was awarded the Excellence in Computer Science award at the 2016 NASA Goddard Poster Fair for this project.