Post-doctoral Position in Machine Learning for Subsurface Multiscale Structure and Characterization

Post-doctoral Position in Machine Learning for Subsurface Multiscale Structure and Characterization including Permeability at the University of Pittsburgh Geology and Environmental Sciences - Pennsylvania-Pittsburgh - (26005201) The University of Pittsburgh is seeking a highly motivated and creative Postdoctoral Researcher to join a cutting-edge project focused on applying artificial intelligence and machine learning (AI/ML) to critical subsurface energy challenges for a three-year post-doctoral appointment. This position is funded by the United States Department of Energy Science-informed Machine Learning to Accelerate Real Time subsurface decision making (SMART) LDRD Prime initiative. The successful candidate will be central to a project aiming to develop a breakthrough, laboratory-calibrated AI/ML tool that accurately estimates subsurface permeability from commonly collected geophysical well logs. This research will address a key challenge in subsurface characterization for applications including energy resources, production efficiency, and recovery optimization. The researcher will work with a multidisciplinary team to build, train, and validate novel deep learning models, leveraging unique datasets from national laboratories. Key Responsibilities The postdoctoral researcher will be integral to achieving the project's ambitious goals and will be expected to: Demonstrated experience with developing relational databases and database schemas. Lead the construction of a fully attributed, machine learning-ready petrophysical database from existing NETL ultrasonic and core measurement archives. Develop, train, and deploy deep learning models, including convolutional neural networks (CNNs) and physics-informed neural networks (PINN), to predict rock permeability from ultrasonic acoustic measurements. Adapt and retrain existing deep learning frameworks (e.g., PhaseNet) to automate the picking of P and S wave arrivals from ultrasonic waveform data, enhancing the speed and consistency of laboratory analysis. Develop and apply generative adversarial networks (GANs) to produce realistic synthetic core data, broadening the training datasets for more robust AI/ML models. Integrate and validate the developed models by applying them to existing wireline log data and potentially new core samples. Collaborate closely with NETL scientists and researchers in geophysics, geology, engineering, and computer science. Publish research findings in high-impact, peer-reviewed View the full job description on the employer's website. Equal employment opportunity, including veterans and individuals with disabilities. PI286684060 / 25-9044.00 Teaching Assistants, Postsecondary