- Bangalore, India
- Senior
- 5+ years
Lead design and optimization of GPU inference stack for large diffusion and transformer workloads.
Responsibilities
- Architect high-performance inference runtimes, kernel dispatchers, and memory planners for large diffusion and transformer workloads.
- Lead investigations into cross-GPU performance bottlenecks, communication overheads, and scheduling inefficiencies.
- Drive multi-GPU parallelism strategies — model, pipeline, and tensor parallelization.
- Establish company-wide GPU optimization standards, tooling, and SLIs.
- Collaborate with research to design scalable implementations of novel architectures.
Requirements
- Nice to have: compiler-aided optimization
- Nice to have: TVM
- Nice to have: XLA
- Nice to have: MLIR
- Nice to have: Stable Diffusion