At Mercor Intelligence and Handshake AI, I evaluated and annotated LLM physics reasoning for top AI labs and wrote JSON rubrics for a SciCode-style benchmark.
At the University of Freiburg, I extended an algorithm for automating perturbative Monte Carlo simulations in Sherpa (C++). This cut higher-order terms by 50%-70%, boosting efficiency of synthetic AI training data generation.
For CERN’s CMS, I computed 1%-5% theory corrections on HPC clusters and advised on incorporating them. I also automated symbolic manipulation of analytic expressions with ∼100k terms in Mathematica.
At the University of Göttingen, I deployed simulations across distributed HPC clusters and processed ∼100GB of data with parallel Python scripts. At CERN, I implemented C++ algorithms that improved signal extraction in large datasets by up to 180%.
