Hi, I'm Jacob Berger

I'm a PhD student in Artificial Intelligence at Oregon State University, working under Dr. Lizhong Chen at the STAR Lab. My research so far has been in LLMs, where I developed Parsimonious Decoding and the Vocabulary Efficiency Score for natural-language semantic constraints; I'm now developing machine-learning methods for preconditioner selection in parametric PDE solvers, with a focus on transport equations, alongside collaborators at CARRE and Lawrence Livermore National Laboratory. I'm also a technical contributor and cofounder at Lexel Synergetics, where I led the technical proposal for our $305,000 NSF Phase I SBIR grant and build LLM pipelines from local GPUs to AWS, GCP, and Vercel. I hold dual B.S. degrees (a double major) in Applied Computer Science and Mathematics from Oregon State University (Summa Cum Laude) and work day to day in Python, PyTorch, and Linux on LLM fine-tuning and serving. I care about work that's rigorous and actually useful.