Co-founder and Chief Scientific Officer at SuperMatics, accelerating superconductor R&D with quantum-mechanical simulation, machine learning and real laboratory data. Previously at Rigetti Computing and Berkeley Lab. Based in Berkeley, California.
I'm a generalist drawn to problems at the intersection of machine learning, scientific computing, and the physical world: how models learn, where they fail, and how to turn their predictions into useful decisions.
- Learning and generative models: optimization, representations, and the numerical details that make training and inference behave as intended.
- Reliable AI systems: LLM serving, tool use, and evaluation, with a focus on correctness across streaming, batching, and persistent state.
- AI for discovery: combining simulation, uncertainty, and experimental data to decide what to measure, test, or build next.
- Computation meets physics: superconducting materials, quantum hardware, and cryogenics, where computational models have to respect physical constraints.
Selected correctness and reliability work across AI/ML, generative media, and scientific computing.
- Start from issues a maintainer has already accepted, and check nobody else is on them.
- Reproduce first. Add a test that fails without the fix, then make the smallest change that passes it.
- Show before and after evidence, run the project's own checks, and report honestly which ones fail and why.
- I use AI coding tools for research and drafting. I review every change myself and disclose AI assistance in each pull request, following each project's policy.