Sambhav worked on expanding tensor operation interoperability in the llvm/torch-mlir repository by enabling TOSA conversions within the Bazel build rules. This feature broadens the range of supported tensor operations, allowing Torch-MLIR to integrate more seamlessly with TOSA-backed backends and production pipelines. Sambhav’s approach involved customizing Bazel build configurations and applying knowledge of C++ and MLIR to orchestrate the conversion workflow. The work addressed integration friction for downstream tooling and laid a foundation for future backend optimizations. While the scope was focused, the implementation demonstrated depth in build system configuration and cross-stack interoperability for machine learning compiler infrastructure.