During a special session on June 10th at the Unconventional Resource Technology Conference (URTEC) in Houston, panelists discussed the latest advancements in unconventional AI and machine learning.
AI – It's not just “easy” solutions. Session participants were first heard from Nuny Rincones, Reservoir Engineering Innovation Manager at Conoco Phillips. For the past three years, Rincones said she has led a team focused on reservoir analysis, probabilistic prediction and machine learning tactics.
“Using technology just to embrace it is not the answer,” warned Lincones. While she praised the benefits of AI applications in the field, she also emphasized the importance of truly understanding workflows and applying AI where it affects most.
“If you have more than 40,000 wells in the Permian Basin, tools like cloud computing are useful,” Rincons said. “We use AI everywhere.”
Machine learning methods. Next was Dr. Uchenna Odi, Petroleum Engineering Specialist, AI Team Leader and Digital Transformation Team Leader at ARAMCO Americas.
ODI gave a detailed presentation of his experiences on the research and use of automated machine learning of phase prediction scenarios.
Phase prediction involves using data from well logs and other geophysical measurements to determine the rock type (or phase) present at a particular location. This process is important for understanding the properties of underground and reservoirs and optimizing hydrocarbon E&P.
“Early in my career, we were trying to better understand drilling through the facies,” ODI explained.
Comparing traditional approaches with machine learning, we realized that ODI can achieve significant time savings.
“The traditional methods for phase prediction and characterization took several months,” ODI said. “Traditional machine learning can reduce that time to weeks, while automated machine learning can reduce study time in just a few days.”
Five years ago, ODI said he left to learn more about the types of startups that have come out of Silicon Valley and how they can be applied to the oil and gas realm. He then developed AI algorithms for data modeling using the Datarobot platform, and then implemented Automated Machine Learning (Automl).
From the lab to the field. The third panelist, Yitian Xiao, investigated similar innovations. Currently at the Deep-Time Digital Earth International Research Center, Xiao previously spent 25 years as a data scientist at Exxonmobil and several years at Sinopec's Petroleum Exploration & Production Research Institute.
In 2025, Xiao said he moved to the lab to work with machine learning models and Genai only. “The energy industry has great potential to utilize advanced AI tools, which could fundamentally change the way data is analyzed.”
Xiao is a senior consultant for the Geogpt project, developed at Zhijiang Lab in China. Geogpt is a non-commercial, open source large model (LLM) used in geoscience. While still in prototype mode, the team hopes that GeoGPT will be released later this year.
Unlock shale and tight oil with AI. To close the session, Travis Clark, a data scientist at Chevron, was exposed to several tools Chevron is working to maximize the value of its oil fields. “We have quite a bit of data from the Permian era,” Clark said.
Chevron is one of the top Permian producers and is one of many oil majors that integrate AI into its operations. Efficiency is more important than ever, as lower oil prices creates difficult break-even prices for projects.
Clark has presented several tools being developed by Chevron, including lateral influx that function to improve well production and FRAC design, and the Simops schedule optimizer “predicts and minimizes potential FRAC hits,” Clark said.
