Senior/Staff Machine Learning Engineer

Job Description

Terra AI is building a new category at the intersection of artificial intelligence, geoscience, and critical resource development.

As global demand for copper, lithium, nickel, rare earth elements, geothermal energy, and other strategic resources accelerates, the mining and subsurface industries face a growing challenge: traditional exploration methods remain slow, expensive, and highly uncertain. Terra AI was founded to help solve this problem by redefining how critical resources are discovered, evaluated, and developed.

By combining advanced machine learning, probabilistic modeling, and deep geoscience expertise, Terra AI helps exploration and mining companies make faster, more informed subsurface decisions with greater confidence and capital efficiency. The company’s platform integrates geological, geophysical, and drilling data into intelligent systems designed to improve targeting accuracy, accelerate discovery timelines, and reduce exploration risk.

Backed by Khosla Ventures, BHP, and Rio Tinto, Terra AI works with industry leaders to bring a more informed, data-driven approach to critical resource development.

Terra AI’s mission is to define the new global standard for data-driven critical resource development — breaking the cost and time curve required to support electrification, energy security, and the global energy transition.

The company operates with a strong partnership mentality, combining technical rigor, candid communication, continual learning, and environmental stewardship to help modern exploration teams solve some of the world’s most important resource challenges.

Who you are

You’re a deep learning engineer or scientist who loves building new models, not just applying existing ones. You’re comfortable working from first principles, experimenting with architectures, training large models, and learning from results that don’t always behave as expected.

You bring deep PyTorch expertise, strong software engineering habits, and the judgment to turn research ideas into reliable experiments. You care about the quality of training data as much as the sophistication of the model, and you enjoy collaborating with people whose expertise extends beyond machine learning.

You’re excited by the opportunity to apply generative AI to a genuinely difficult physical-world problem: understanding the geology we cannot directly see. Prior geoscience experience is welcome, but curiosity and a willingness to learn matter just as much.

What you’ll do

Terra AI is developing generative models that produce three-dimensional geological representations conditioned on geophysical surveys, borehole measurements, and other physical observations. Like image-generation systems that produce different plausible images from a prompt, our models aim to produce plausible views of the subsurface consistent with the evidence available.

The goal is not to pretend we know exactly what lies underground. It is to capture both what we know and what remains uncertain, giving exploration teams a stronger foundation for deciding where and how to investigate critical resources.

In this role, you’ll lead the development of these models from architecture and training through evaluation and adaptation to real projects. You’ll work closely with other AI researchers, engineers, geoscientists, and project teams to connect novel modeling approaches with consequential real-world decisions.

Core responsibilities

• Design, implement, train, evaluate, and iterate on diffusion models that generate 3D geological models.
• Develop approaches to condition model outputs on geophysical surveys, borehole measurements, and other physical observations.
• Shape synthetic-data generation and data-curation approaches to improve model quality, robustness, and generalization.
• Establish rigorous experiments, evaluation methods, and reproducible training workflows to understand model performance and limitations.
• Adapt generative modeling approaches to real-world projects in close collaboration with geoscience and delivery teams.
• Build maintainable, well-tested model code and contribute to the engineering foundations needed for continued experimentation and deployment.

Qualifications

Required

• Extensive hands-on PyTorch experience, including custom modules, training optimization, and debugging complex or large-scale models.
• Demonstrated experience designing, implementing, and training substantial deep learning architectures from scratch.
• Strong data-curation skills, including creating, cleaning, validating, and maintaining datasets for machine learning.
• Solid software engineering and system-design fundamentals, including version control, testing, code optimization, scalability, and maintainability.

Nice to have

• Experience with generative architectures, especially diffusion models and posterior sampling methods.
• Experience developing and training transformer architectures, particularly for 3D or other structured data.
• Experience scaling model training across multi-GPU systems or large GPU clusters, including distributed training optimization.
• Experience managing and optimizing cloud infrastructure for machine learning workloads.
• We recognize that exceptional candidates come from a range of research, engineering, and scientific backgrounds. If this work excites you and your experience aligns with the spirit of the role, we encourage you to apply even if you do not meet every preferred qualification.

Compensation & Benefits

The annual base salary range for this position is $255,000-275,000. Actual compensation will depend on experience, qualifications, skills, location, and other job-related factors. Any additional incentive compensation will be communicated separately. Applications may be submitted through the job posting; the application closing date will be provided when the role is published.

We’re committed to building a workplace where people with different backgrounds, experiences, and perspectives can do their best work. If you need an accommodation during the application or interview process, please let us know.