Steven Kolawole

I am a second-year PhD student at Carnegie Mellon University, specializing in efficiency challenges in machine learning. My research focuses on optimizing ML inference through model routing, architecture design, and execution efficiency. I have developed and evaluated methods such as Agreement-Based Cascading (ABC), which routes tasks to smaller models when possible, reducing costs for edge-to-cloud and LLM API setups, and Bonsai, an inference-time pruning framework that creates smaller, faster models for constrained hardware.

I’ve presented my work at conferences, published research papers on ML efficiency, and contributed to open-source repositories on inference optimization.