GPU Compute & MLIR Compiler Engineer — AI Workloads

Overview

GPU Compute & MLIR Compiler Engineer focused on designing, optimizing, and deploying high-performance GPU compute kernels and MLIR compiler backends for AI workloads.

What you'll do

• Develop and optimize GPU compute kernels targeting OpenCL and Vulkan for high-throughput AI/ML.
• Design and extend MLIR dialects across frontend, graph-level IR, tensor IR, and runtime/low-level layers.
• Implement MLIR compiler passes for tiling, fusion, bufferization, vectorization, and lowering pipelines.
• Profile compiled kernels using GPU counters/vendor profilers and drive compiler-level performance improvements.
• Build and maintain GPU runtime infrastructure for OpenCL and Vulkan (memory management, pipeline setup, command orchestration, scheduling).
• Collaborate with ML teams to optimize end-to-end model lowering for computer vision and LLMs.

What you'll need

• Hands-on experience authoring/extending MLIR custom dialects, writing compiler passes, and building end-to-end lowering pipelines.
• Deep MLIR expertise across frontend dialects (e.g., TOSA/StableHLO/ONNX-MLIR), graph-level IR, and tensor IR.
• Knowledge of runtime/low-level dialects including bufferization, MemRef, SCF, GPU, and LLVM dialects.
• Hands-on OpenCL experience including kernel development and OpenCL runtime management.
• Solid understanding of Vulkan compute programming and synchronization primitives.
• Proficiency in C/C++ for system-level development.

• Experience with GPU kernel profiling and bottleneck analysis.

Details

• Location: Bangalore, India.