Member of Technical Staff - Research, Physics

About the role

Causal is developing a Large Physics foundation Model (LPM) to understand cause and effect in physical systems, starting with weather prediction. This role involves integrating deep physics knowledge into our AI models to ensure they accurately reflect physical laws and generalize across different domains. You will contribute to building AI capable of predicting and influencing future outcomes. The position requires a deep thinker who can bridge abstract physical theory with practical engineering constraints. You will own the physical validity of the models throughout their lifecycle. Your work will directly determine whether the system respects fundamental laws of nature. This is a chance to define a new discipline where physics and machine learning are inseparable.

Key facts

What you'll do

• Translate high-level research objectives into concrete physical constraints that can be enforced within the Large Physics foundation Model.
• Design and execute rigorous physical consistency tests that probe conservation laws, symmetry principles, and domain-specific invariances.
• Partner with data teams to audit training data for violations of thermodynamic principles and numerical stability issues.
• Evaluate the LPM across diverse physical regimes to map its domain of reliable operation and identify failure modes.
• Collaborate with model architects to embed numerical solvers that respect discretization constraints and stability criteria.
• Lead reviews of model outputs to flag unphysical predictions that could mislead downstream decision-making processes.
• Work closely with interpretability researchers to surface the physical reasoning pathways used by the foundation model.
• Define benchmarks that quantify how well the model extrapolates to unseen physical scenarios without violating first principles.
• Act as a physics authority in cross-functional discussions, shaping product requirements based on scientific rigor.
• Document physical assumptions and their implications for model behavior in a clear, accessible format for engineers and scientists.
• Apply physical principles to the model, evaluating its consistency with conservation laws and physical constraints.
• Create assessments to verify the model's physical coherence, beyond just statistical accuracy.
• Offer expertise on the physics of systems we model, including fluid dynamics and thermodynamics, and their numerical handling.
• Examine the LPM's ability to generalize across various physical domains and identify its limitations.
• Work with model, evaluation, and interpretability teams to align physical understanding with research goals.

Requirements

• Hold a PhD or possess equivalent research experience in physics, with a strong focus on areas such as fluid dynamics, thermodynamics, or computational physics.
• Demonstrate extensive experience with numerical simulation of physical systems, including Computational Fluid Dynamics (CFD) and an understanding of the modeling trade-offs involved.
• Show a sustained interest in the intersection of machine learning and physical modeling, evidenced by publications or projects in related domains.
• Possess a proven ability to work effectively with machine learning researchers and to translate complex physics concepts into precise technical specifications.
• Apply a precise, data-driven methodology for assessing model quality, using quantitative metrics grounded in physical laws.
• Have hands-on experience with at least one major programming language commonly used in scientific computing and machine learning workflows.
• Be comfortable operating in an environment where research questions are tied to real-world impact and deployment constraints.
• Exhibit strong analytical and problem-solving skills, with the patience to debug issues that span multiple layers of the modeling stack.

Skills & tools

• Fluid dynamics
• Thermodynamics
• Computational physics
• Numerical simulation
• CFD
• Machine learning
• Physical modeling

Practical notes

This role is based in San Francisco and requires full-time on-site engagement. The position is not eligible for remote work or relocation assistance. Candidates must be authorized to work in the United States without sponsorship for this role. No specific travel requirements are outlined, but participation in internal reviews and cross-team workshops is expected. There are no published deadlines for this posting, but early submission of materials is encouraged due to the specialized nature of the role.