Vijay Suresh developed robust Welford-based statistical capabilities for the tenstorrent/tt-metal and tenstorrent/tt-llk repositories, focusing on accurate multi-stream mean and variance calculations within performance-critical machine learning kernels. He unified the Welford statistics API, enabling consistent handling of statistical updates across architectures, and integrated this into kernel computations for both LLK LayerNorm and Blackhole architectures. Working primarily in C and C++, Vijay emphasized low-level kernel development and numerical algorithm implementation, ensuring improved numerical stability and performance. His work established a foundation for reliable streaming analytics, demonstrating depth in embedded systems and hardware acceleration through careful cross-architecture integration and testing.