Rafael Vasquez

Rafael Vasquez is an Open Source Software Developer at IBM in Toronto, Ontario, where he most recently focused on contributing to Spyre's vLLM plugin and inferencing stack. He previously helped maintain KServe/ModelMesh, a multi-model serving platform for Kubernetes. His interests span inference optimization, model serving architecture, and hardware-aware deployment of foundation models.

Rafael holds a Master of Applied Science from Toronto Metropolitan University (formerly Ryerson University), where his research applied deep reinforcement learning to autonomous vehicle safety. He authored a conference paper on training an autonomous braking system, and co-authored published work on drivers' willingness to ride as passengers in autonomous vehicles, and developed novel methods for pedestrian intention and path prediction. He also led development of an extensible virtual reality sandbox built in Unity Engine for designing interactive transportation experiments, integrating traffic simulation, driving hardware, and data collection.

Before returning to IBM, Rafael worked as a Data Scientist at Loblaw Companies Limited, where he built algorithms to detect declining patterns in grocery customers' transactional behaviour and engineered customer risk scores used to drive personalization across their loyalty points program.

Rafael is active in the open source AI community, having presented internationally at Open Source Summit North America and Japan, AI.dev North America and Europe, ODSC East, and DevOpsDays Geneva. He also mentors new contributors, authors technical tutorials, and organizes workshops on technologies including Granite and watsonx.