tardigrade-cad

Physics-constrained CAD for magnet and coil design

Overview

JAX-based Biot–Savart field solver with optax-driven geometry optimization for designing coils and magnet assemblies against physical field constraints, not just geometric ones.

Architecture

• JAX-based Biot–Savart field solver as the core physics engine
• optax-driven gradient-based geometry optimization against explicit field constraints, not just geometric tolerances
• Built and validated; used for coil and magnet assembly design

Milestones

1. 01Validate the solver against known analytic coil configurations (e.g. a Helmholtz pair) as a standing regression test
2. 02Package for reuse across other hardware programs — the Earth-field NMR polarizing coil is a direct candidate
3. 03Publish to PyPI alongside tardigrade_agent

Open questions

• Whether to extend beyond magnetostatics to thermal or mechanical constraints
• How tightly to couple this into the NMR qubit and MPCVD hardware programs versus keeping it a general-purpose tool