University of Michigan professor with expertise in AI reliability, SQL and query optimization, database performance, cloud data warehouses, and FinOps.
Barzan Mozafari is a University of Michigan professor, systems researcher, and founder and former CEO of Keebo. His experience spans AI reliability, database systems, query optimization, cloud data warehouses, and FinOps. Across academic research, open-source contributions, commercial products, and expert witness engagements, his work has consistently centered on technically demanding problems at the intersection of AI, data, and cloud infrastructure.
Whether an AI system has met a user’s intent is a different question from whether it has produced a plausible result. His current research examines incomplete specifications, incorrect assumptions, and ways to reduce the verification burden placed on users.
Research →His work covers SQL rewriting and equivalence, query execution, transaction scheduling, physical design, and performance diagnosis. Representative projects include GenRewrite, DAGSmith, DBSeer, and DBSherlock.
Research →Cloud performance and cost depend on the interaction between workloads, resource allocation, and execution. His work in research and as Keebo’s founder and CEO has addressed warehouse sizing, workload-aware optimization, and performance–cost tradeoffs.
Research →