Rafid Mahmood
I am an Assistant Professor at the Telfer School of Management in the University of Ottawa. I am also a part-time Sr. Research Scientist at the NVIDIA Toronto AI Lab.
I am interested in the operational challenges behind the deployment and use of AI systems. I use deep learning and data-driven optimization for problems where the typical forms of large-scale data collection and model tuning are prohibitive. My work addresses ML technology (e.g., computer vision systems), healthcare (e.g., personalized medicine), and finance (e.g., portfolio optimization). I am also interested in general data science problems (e.g., sports analytics).
Students interested in (Undergraduate/Masters/PhD) supervision at Telfer are welcome to contact me here with a CV and transcript.
Graduate students interested in internships at NVIDIA are welcome to contact me here with a CV and summary of research interests.
News
| Apr 15, 2025 | Check out our pre-print, Minority Reports: Balancing cost and quality in ground truth data annotation. We collaborate with Quality Match to better manage crowdwork data annotators. |
|---|---|
| Mar 8, 2025 | Our paper Optimizing Data Collection for Machine Learning was accepted to the Journal of Machine Learning Research (JMLR). |
| Feb 26, 2025 | We are organizing the Exploring the Next Generation of Data workshop at CVPR 2025! The paper submission is now open. |
| Feb 26, 2025 | Our paper Can large Vision-Language Models correct grounding errors by themselves? was accepted at CVPR 2025. |
| Sep 28, 2024 | Our paper, Pricing and Competition with Generative AI, was accepted at NeurIPS 2024. See you in Vancouver! |
| Sep 20, 2024 | Our paper, Reasoning Paths with Reference Objects Elicit Quantitative Spatial Reasoning in Large Vision-Language Models was accepted at EMNLP 2024. |
| Jul 29, 2024 | Our pre-print, AutoScale: Automatic Prediction of Compute-optimal Data Composition for Training LLMs, introduces a method for determining the optimal mixture of data to train LLMs. |
| Apr 9, 2024 | Check out our pre-print, Can Feedback Enhance Semantic Grounding in Large-Scale Vision Language Models, which uses multiple VLMs that iterately improve semantic prediction tasks! |