Global energy / commodity trader


A large global vertically integrated energy company sought to understand the destination of ships carrying LNG cargo within an hour of departure, to better understand market dynamics The asked Faculty to build a machine learning model to make predictions and deploy the solution on their AWS infrastructure.


We built a solution that incorporated a new LNG shipping database, making the data more usable and accessible across trading desks. The model takes in live data on a ship’s location (pings) and makes a prediction on its destination one hour after leaving the port which is updated every 5 minutes. Analysts can query trades within a date range and compare live predictions against historical data.


Our platform outperformed existing industry software with 76% accuracy. It also provides features not commercially available such as longer timeframes for predictions, more frequent updates and probabilities associated with destinations.