Masoud Rahimi

Biography

Masoud Rahimi is a PhD candidate at LECS, working on accelerators that process large sparse graphs directly inside resistive memory. His research targets the data movement that dominates graph analytics: carrying the graph in its compact coordinate list form from main memory into the crossbars so that no format conversion is needed, and reusing crossbar configurations across the connection patterns that recur throughout real-world graphs. Recent contributions include a conversion-free graph accelerator built on a unified sparse representation (IEEE Transactions on Emerging Topics in Computing, 2026), canonical pattern normalization for pattern reuse in ReRAM crossbars (IEEE NEWCAS 2026, First Place Best Paper Award), static engine assignment for recurring subgraph patterns (DATE 2026), and massively parallel hyperdimensional computing architectures for intelligent satellite links (ACM Computing Frontiers 2026). His current directions are on-chip communication for larger accelerator arrays and hardware techniques that keep in-memory computation correct as resistive cells age.