Chinmay Dattanand Kuchinad

Senior Deep Learning Engineer - Ml Frameworks (Pytorch) @AMD

Austin, TX, US

Implemented multi-architecture compilation in PyTorch Inductor for ROCm, designing a HIP-based code-generation and linking flow that allows Ahead-of-Time (AOT)–compiled artifacts to execute across heterogeneous AMD GPU architectures- Ported the StaticCudaLauncher runtime to ROCm, enabling static kernel launching and direct.hsaco binary execution within Inductor’s runtime launcher, bringing feature parity with CUDA’s.cubin pipeline- Extended Inductor’s memory-management layer to support dynamic storage-resize operations on ROCm devices, unifying CUDA-specific resize_storage_bytes logic under a cross-backend interface used by FSDP + Inductor distributed training- Collaborated with Meta’s PyTorch Inductor and Triton compiler teams to define the ROCm equivalents of PTX emission and fat-binary packaging, contributing design feedback on multi-arch build strategies and HIP CMake integration- Enhanced ROCm compiler integration and debugging workflows, introducing architecture-aware build flags, improved HIP diagnostics, and streamlined CI coverage for Inductor kernel generation across MI200/MI300 GPUs

PES University

Bachelor of Technology - BTech, Electrical, Electronics and Communications Engineering

University of Southern California

Master of Science - MS, Electrical Engineering

Gokhale Centenary College, Ankola.

PU, Science

BGS Central School,Mirjan

High school