Ronnie

I’m particularly interested in local LLMs, model behavior, and inference optimization. My work focuses on running and evaluating open-weight models on Apple Silicon, exploring quantization strategies, prompt engineering, and long-context performance. I enjoy experimenting with techniques such as model ablation, behavioral alignment, and custom evaluation to improve instruction following, creative writing, and narrative consistency. More broadly, I’m interested in understanding how internal model behavior can be shaped beyond traditional fine-tuning, especially for specialized creative and reasoning tasks.