Senior/Staff Machine Learning Researcher

Role Description

The company is developing a generative model that produces 3D geological models conditioned on geophysical surveys, borehole measurements, and other forms of physical observation. This model aims to revolutionize decision-making in the earth subsurface for a wide range of clean energy applications.

Role Responsibilities

• Design, train, test, and iterate on diffusion models for 3D geological models
• Design, train, test, and iterate on an approach for conditioning generation on geophysical data and other observations
• Inform the generation of synthetic data to improve model performance
• Adapt diffusion modeling approach to specific real-world projects in collaboration with project teams

Qualifications

• Extensive PyTorch Experience
• Deep understanding of PyTorch, including writing custom modules, optimizing training, and debugging issues in large-scale models
• Expertise in Developing Large Deep Learning Models from Scratch
• Proven ability to design, implement, and train complex deep learning architectures from the ground up
• Data Curation Skills
• Hands-on experience in creating, cleaning, and maintaining high-quality datasets tailored for machine learning applications
• Strong Software Engineering and Design Experience
• Proficient in software development best practices, including version control, testing, and code optimization
• Familiarity with designing scalable and maintainable systems

Nice-to-haves

• Experience with Generative Models
• Familiarity with generative architectures, particularly diffusion models, and an emphasis on posterior sampling methods
• Knowledge of Transformer Architectures
• Experience building and training transformers, especially in applications involving 3D data
• Scaling Models Across Large GPU Clusters
• Expertise in parallelizing models across multiple GPUs and optimizing distributed training pipelines
• Cloud Infrastructure Expertise
• Experience setting up, managing, and optimizing cloud environments for machine learning workloads, including provisioning resources and managing costs