Finding oil and gas using DNA from soil samples.

Results are encouraging.

Client: Biodentify

Analysis and further testing

Industry: Oil and Gas

Project Duration: 12 months

Goal: Optimize and automate the prediction of oil and gas underground

Tech: Reinforcement learning, Keras, Tensorflow, Scikit-learn

A positive result

The challenge:

Soil is sampled and DNA is extracted

Many DNA features (>330k) and few samples (<3k)

Predict with at least 70% accuracy whether there is gas or oil under the ground

The approach:

Automate A.I. procedures using Automated Machine Learning

Create new features and automatically test them

Robolect proposed solutions that the company had not thought possible

The solution:

We created a reinforcement learning agent for the Automated Machine Learning to generalize on new data

The results:

More than 70% correct results, as the customer wants