Data Scientist and Developer in Denver, CO, United States

This was a strategy that could be used to market make highly liquid commodities and futures markets. It focused on market microstructure to inform the model and make forecasts. I was in charge of building and testing the model that was used to predict price movements over the next second. The general framework for this strategy was used to create models for over 20 different markets. In its more than six years in production, it consistently produced annual returns greater than 75%.

The goal of this project was to accurately predict day ahead and real time prices in various power markets. Prices were predicted one day to one week into the future. It employed machine learning techniques taking into account the temporal nature of power data. Data was aggregated from various sources (ISOs, weather data, ICE market data) to create the final model. The project led to a pricing model that accurately forecasted market prices, producing annual returns of around 200%.

The project involved creating a script to automatically classify load forecasts for various power markets. Load forecasting is an integral park of any price model in power markets, and as such, correctly classifying the type of curve is crucial. Accurately classifying the shape of the curve was done to remove any human subjectivity and aid in automating manual processes.

This was a web page that provided on-demand reporting about all current and past trading strategies. I was in charge of building and maintaining this site. The site was built using the R-Shiny platform. It provided users a quick look at any trading strategy over the previous six months. Through a series of widgets and dashboards, it allowed users to quickly monitor business and trade performance and make data driven decisions going forward.