Technical founder working across ML systems, scientific computing, and forward-deployed AI.

I founded Mirror Progress, where I have led in enterprise technical delivery from stakeholder discovery and architecture through implementation and deployment. My independent research and engineering work focuses on model training and evaluation, state-aware systems, scientific ML, and the infrastructure connecting models to real organizational workflows.

Ronnie's Lab | UC Berkeley research profile | LinkedIn | Email

PyTorch transformer and hybrid state-space model systems, Slurm-backed execution, checkpoint lineage, model evaluation, FastAPI services, Next.js observability, and scientific-ML experiments.

A governed, state-aware computing program connecting numerical execution, evidence, permissions, fallback, and agent decisions across heterogeneous systems.

A bounded execution methodology for exact modeled reuse, compact counterfactuals, complete affected-closure propagation, and full-replay fallback.

Programmable photonic computing research focused on mixed-signal survivability, readout constraints, resident sparse decisions, and falsifiable architecture design.

• Build below the demo layer.
• Keep model, data, evaluation, and deployment boundaries inspectable.
• Publish corrections and negative results alongside positive evidence.
• Translate technical systems into measurable organizational outcomes.