About Me

My name is Daniel Israel, and I am currently a Computer Science PhD student at UCLA. I am co-advised by Guy Van den Broeck and Aditya Grover. LLMs are continuously being optimized for performance at every layer of the stack: hardware, kernels, architecture, training, and inference algorithms. My research focus is primarily directed at more performant inference algorithms, in some instances enabled by modest architectural modifications. LLM inference algorithms can be improved by

  1. Removing redundant computation and providing an exact solution that is strictly better
  2. Identifying a tradeoff and providing an approximate solution that expands the Pareto frontier of speed and quality

In my work, I have employed both methods towards faster LLM inference algorithms. Before starting my PhD, I am thankful to have worked with professors Anima Anandkumar, Frederick Eberhardt, and Pietro Perona at Caltech. In my spare time, I enjoy playing soccer and chess.