• Alternative data

• Alternative data

• Created end-to-end solution to predict retailers' quarterly earnings through interpretation of

satellite imagery.

• Implemented and trained M-RCNN to identify cars from satellite imagery with 92% accuracy.

• Developed regression model to correlate results from M-RCNN to quarterly sales. Predicted

revenue fell within 12% of true revenue.

• Designed and constructed automated data pipeline, transforming county-wide MrSID files into JPEGs containing a store's parking lot.

• Developed cloud-based malware scanner hosted on Google Cloud Platform

• Helped develop inference engine and corresponding classification model using TensorFlow and other M.L. libraries.

• Incorporated a micro-services based application architecture.

• In-application database of malware corpora updated in real-time with results of scanner.

• Reduced and simplified retraining methodology to simply pushing new batches of malware

corpora.