Sandisk

Staff Test Engineer - AI SW Stack ( 9-13 years)

Sandisk
Apply
16 days ago
Bengaluru, IndiaStaff+

Responsibilities

  • Define and own end-to-end validation strategy for the AI software stack, including methodologies, coverage metrics, quality gates, and release-readiness criteria.
  • Lead validation of neural-network operators, compute kernels, graph-level execution, end-to-end model inference, and software behavior across hardware revisions.
  • Validate memory allocation and movement across host and device environments, including stress, concurrency, and multi-device scenarios.
  • Design and maintain benchmarking frameworks for operator performance, model performance, latency, throughput, time-to-first-token or result, and multi-device scaling efficiency.
  • Mentor engineers and drive automation, performance-validation, debugging, and software-quality best practices.
  • Lead cross-functional validation efforts with compiler, runtime, kernel, and hardware teams.

Requirements

  • 10+ years of experience in AI software validation, systems software validation, performance engineering, or related areas.
  • Strong C++ software development expertise and object-oriented programming fundamentals.
  • Advanced Python programming skills for automation and test infrastructure.
  • Deep understanding of AI/ML execution pipelines, neural-network operators, deep-learning frameworks, runtime architectures, scheduling systems, parallel execution models, memory hierarchy, and DMA concepts.
  • Experience validating large-scale AI workloads and inference systems.
  • Proven experience building automated test and benchmarking frameworks, with strong debugging and root-cause analysis skills.
  • Preferred: experience with MLIR, XLA, StableHLO, TVM, ONNX Runtime, TensorRT, or similar compiler/runtime stacks.
  • Preferred: experience validating GPU, NPU, FPGA, or custom accelerator platforms.
  • Preferred: hands-on experience with VTune, Nsight Systems, Nsight Compute, perf, or gprof.
  • Preferred: experience with distributed inference and multi-device execution, familiarity with LLM, multimodal, and generative AI workloads, and experience leading software quality or validation teams.

Tech Stack

Categories

Sandisk

About Sandisk

5,001-10,000 employees
Contact me