I am a PhD researcher in compiler optimization for machine learning systems. My work focuses on graph-level transformations and developing new heuristics for the fusion problem in neural network computational graphs, aiming to reduce memory movement and improve performance across heterogeneous architectures. I work with modern compiler infrastructures such as Google XLA, Apache TVM, and MLIR/LLVM, exploring tiling, bufferization, and hardware-aware optimizations. My research bridges compiler design and High-Level Synthesis (HLS), contributing to more efficient, memory-conscious compilation pipelines for deep learning workloads.
