Nouman Amir developed the Minimum operation (MinOp) for quantized LLM and GenAI workloads in the iree-org/wave repository, focusing on efficient element-wise minimum computations within the Tensor Kernel Wave (TKW) library. He implemented the lowering of the min operation to floating-point, signed, and unsigned integer arithmetic, updating both the Python interface and the TKW_COMBINER decomposition logic to support the new functionality. Nouman also created comprehensive end-to-end tests to ensure correctness across data types and shapes. His work in compiler development and low-level optimization improved performance and latency for GenAI inference on quantized models, demonstrating strong technical depth.