Summary
Mishaal Kazmi is a Machine Learning Researcher with eight years of engineering and research experience, currently leading privacy auditing and federated learning projects at the University of British Columbia. He combines systems-level expertise in distributed ML pipelines and HPC (CUDA, Slurm) with practical privacy techniques such as differential privacy and membership inference attacks, and co-developed PANORAMIA—a post-hoc privacy auditing framework published at NeurIPS 2024. His work spans vision and language foundation models (including large CLIP models on 400M image-text pairs), synthetic-data-driven healthcare applications, and scalable DP fine-tuning using Opacus. Previously he bridged product and engineering at Educative, scaling platforms and shipping ML-driven features, and he has a strong track record teaching and mentoring in distributed systems and AI. Notably, he has operationalized large-scale privacy evaluations without retraining models, revealing real vulnerabilities in foundation models while improving the efficiency of attack pipelines.
8 years of coding experience
4 years of employment as a software developer
BS honors, Computer Science, BS honors, Computer Science at Lahore University of Management Sciences
Master's degree, Computer Science, Master's degree, Computer Science at The University of British Columbia
High School, High School at SICAS
A-Levels, 90.3%, A-Levels, 90.3% at LGS
English, Urdu, German