A DESTINED BEGINNING
Our involvement in data science was not intentional. We were toiling away at our respective companies, leading teams searching for methods to accurately predict power plant output and to understand the drivers of shale gas reservoirs. Our physics-based models were failing us and we were frustrated. It was 2012 and we could not find a way to model the interdependencies, non-linear relationships and feedback loops in our respective processes.
Fortunately, we were introduced to machine learning and mathematical optimization when all looked lost. Over a weekend, we put together our first rudimentary models. Despite our lack of sophistication, the models were much more accurate than our previous efforts. We had found the answer. In 2014, we
first met while pursuing our Master’s in Data Science. We formed a lifelong bond learning how to build models to solve tough industrial problems. After school, we went our separate ways, leading data science teams for two of the largest energy companies in the US. We built models that covered a wide
range of processes and created over $200 million in value for our respective companies.
PAYING IT FORWARD
After six years of building enterprise-scale models and applications, making plenty of mistakes and learning many lessons, we yearned for something more – to introduce data science techniques to other companies. We began by hosting meet-ups and seminars to teach others what we had learned about
the practical implementation of methods such as anomaly detection, recurrent neural networks and enterprise scale applications. But it wasn’t enough. So, we joined together to have a bigger impact. Our goal is simple, machine learning for every organization, to help as many people and companies make
better decisions, solve problems and improve their financial outcomes.
SOLVING THE UNSOLVABLE
Over the past six years, we have developed a best practice approach to value creation by building a
wide range of prediction models
OIL & GAS
Forecast production for over 1 million oil and gas wells. Built 3D maps of oil and gas reservoirs. Identified drilling locations with highest production potential. Optimized drilling techniques to maximize rate of penetration and drill bit life, while minimizing cost. Modified completions approach to increase reserve potential. Monitored SCADA data to identify wells offline in real time.
Predicted plant output for fleets of 65 plants, over 21GW and all types of equipment configurations. Forecast power prices in multiple competitive markets. Minimized start-up time to reduce fuel burn while idle. Automated equipment inspection. Identified equipment likely to fail. Predicted equipment mean time to failure.