Projects

My research operationalizes a simple loop: Generation → Execution → Feedback → Training. Each project below tackles a different dimension of this loop—testing and execution feedback, formal verification and proof, and preference-based alignment—composing structured signals into training objectives that push models toward genuine correctness.

Loading projects...

These projects share a common architecture: connect model generation to real-world execution, extract structured feedback from the outcomes, and feed it back into training. Together, they represent different facets of a unified research vision—building models where reliability emerges from the training process itself. Read the full research statement →