NullDeltaQ

Physics-grounded AI — where the Laplacian is stationary.

phase_space.png

What I’m building: AI architectures where physics is the inductive bias — energy-based models, symmetry-aware architectures, geometric deep learning, dynamical systems. Not physics as metaphor. Physics as the actual structure of the model.

NullDeltaQ is the null space of the Laplacian of Q — the harmonic regime where Dirichlet energy vanishes. That’s the mathematical condition we’re building toward.

Background: I spent 16 years leading data science and ML organizations across fintechs financial services, online and data platforms — close enough to production to know what breaks, senior enough to decide what gets built. In 2024 I stepped back from operating roles to focus entirely on the research question I couldn’t stop thinking about: what happens when you build AI from physical first principles rather than statistical ones?

Research focus:

  • Energy-based models and their physical interpretation
  • Symmetry-aware neural architectures (geometric deep learning)
  • Dirichlet energy as structural prior in learning systems
  • Dynamical systems as foundation for inference

LinkedIn · arXiv