For this project, we’re looking for a seasoned Data Scientist with a strong foundation in Computer Vision and/or working experience with ML for imaging and signal processing algorithms. The ideal candidate is an expert problem solver, with a knack for quickly breaking down complex natural science issues that can benefit from nonlinear optimizations. Acute familiarity with seismic processing is extremely beneficial, but not required.

Project details: residual move-out (RMO) picking is a routine, often manual, procedure in seismic depth processing workflows for iterative velocity model building. Based on a supplied velocity model, reconstructed seismic images of the subsurface are spatially re-positioned, or “migrated”. RMO “picks” on 2D velocity semblance panels are used to iteratively refine/update the initial velocity model, which then gets re-migrated with the updated velocities. Analogously, this is not dissimilar to automated image focusing algorithms for DSLR cameras.

Key objectives for this project include:

• Train a deep neural network model to improve RMO picks on seismic data.

• Improved seismic images resulting from an RMO-based model updates, leading to more reliable depths that are crucial for well placement, volume assessment, identifying new hydrocarbon opportunities, etc.

• An automated ML solution is optimal; other solutions meeting objectives are acceptable.

What You’ll Bring:

• 5+ years of AI/ML experience w/strong expertise in neural networks and SOTA models

• Experienced Python programming for scientific computing

• Ability to integrate large, optimized data files (examples: hd5f, pickle, parquet, etc.)

• 3+ years of geophysical/seismic analysis background is beneficial, but not required