Formation Identification

• Used Random Forest to train a model to identify the formations 

• Used 'gamma' data collected from sensors during drilling

• Flattened the data for every inch to build data points for modelling

Text Comment Mining

• Used NLTK for text processing

• Used TF-IDF technique to identify important terms being discussed in the comments

• Identified terms were grouped into bigger categories, which get displayed on the app

License Plate Detection

• Initiated the project to use License plate detection for vehicle tracking

• Used Google Vision API for extracting text off the images

• Sorted mismanaged invoices while saving $10-20K costs/well

OCR Text Detection

• Digitized millions of old paper contracts, invoices using GCP

• Designed a feature similar to google maps ETA for drilling operation (tripping) 
• The estimated time provided app users to better manage their time 

• Used neural networks to classify patterns in rod pump cards
• An early warning saved upto $70K/well with just $100 worth of precaution

• Use attributes of existing wells to recommend reference wells for the new well
• The algorithm provided certain adjustable levers for the end user to suit their needs