Specialized R&D consultancy bridging the gap between rigorous wave physics and modern deep learning. Over a decade of independent expertise in high-performance seismic imaging.
Providing technical leadership and software solutions for complex geophysical problems.
Expertise in 1D, 2D, and 3D propagators including Acoustic, Elastic with topography, and Born modelling. Implementation of advanced gradients like double backpropagation.
Writing high-performance C and CUDA kernels for computationally intensive time loops. Optimizing memory usage across GPU, CPU, and disk for large-scale models.
Pioneering work in differentiable physics. Integrating wave equations into PyTorch graphs for end-to-end training of FWI, RTM, and neural network-based inversion.
A high-performance wave propagation module for PyTorch.
Deepwave enables researchers and engineers to treat wave propagation as a differentiable layer within deep learning pipelines. It combines the ease of use of PyTorch with the raw speed of custom C/CUDA kernels.
Applying active learning to optimally select samples for labelling in first break picking. An approach to assist labour-intensive processing tasks without resorting to black-box neural networks.
Read Article →A U-net constructed with a pretrained ResNet, incorporating information from neighbouring gathers, to denoise and deblend real datasets from diverse geological settings.
Read Article →Integrating a GAN into seismic Full-Waveform Inversion to reduce model parameters and restrict inverted models to plausible geological structures.
Read Article →Demonstrating that conventional FWI algorithms can be constructed as recurrent neural networks, enabling implementation via standard deep learning frameworks.
Read Article →