The Data Guild consulted on the development of a malfunction and theft detection system for energy utility providers. The client’s analytics engine enabled petabyte=scale grid data aggregation of historical and operational data, including smart-meter data, billing and service history, and various metadata.
- Our approach built on work by a team from Berkeley, innovating using a bagging and unsupervised support vector machine (SVM).
- Technologies utilized included python-based libraries (pandas, sklearn, etc) on AWS-based parallel computation.
Aman’s role was as a senior machine learning consultant. I helped design and implement the approach, identifying and ranking potential theft sites.