AI Validation, Workload Enabling and Tools Engineer

Intel Intel · Semiconductors · Bangalore, India

AI Software Solution Engineer focused on validation and workload enabling for Intel platforms. The role involves optimizing AI model efficiency, accuracy, and performance by working with frameworks, algorithms, and hardware. Key responsibilities include enabling AI models on Intel GPUs, debugging deep learning models, conducting benchmarking and validation, developing automation pipelines, and evaluating AI models against competitors. The role also involves customer engagement for enablement and performance improvements, and translating AI workload needs into architecture insights.

What you'd actually do

  1. Collaborate with cross-functional hardware and software engineering teams to validate AI workloads for Intel architectures.
  2. Evaluate and debug deep-learning models, kernels, and operators to maximize performance and efficiency while maintaining accuracy.
  3. Conduct benchmarking, regression analysis, and algorithmic validation across a variety of use-cases and frameworks.
  4. Develop prototype workloads, tools, and automation pipelines to accelerate performance tuning and validation workflows.
  5. Conduct performance and accuracy evaluation of AI models on competition HW to detect and plug the gaps.

Skills

Required

  • Python (NumPy, SciPy, Pandas, PyTest)
  • Linux development/debugging experience (git, cmake, gdb, strace, perf)
  • Git/GitHub/Gerrit workflows and CI/CD automation
  • Docker/Kubernetes, virtualization, performance benchmarking, and automation
  • Strong analytical, problem-solving, and communication skills

Nice to have

  • Working knowledge of Agentic AI deployment
  • Understanding of distributed systems, HPC/GPU scaling, MPI/torchrun/Fully Sharded Data Parallel/Tensor Parallel, and high-performance networking (Ethernet/InfiniBand)

What the JD emphasized

  • Knowledge of ML/DL is a Must including LLM Architecture, Transformer, Attention, Low precision Data type (fp8/fp4), Quantization Techniques, Open source upstreaming, Inference Serving etc.
  • AI Workload enabling including accuracy debug and performance optimization is must.

Other signals

  • Enabling AI models on Intel GPUs for accuracy and optimize for performance
  • Evaluate and debug deep-learning models, kernels, and operators to maximize performance and efficiency while maintaining accuracy
  • Conduct benchmarking, regression analysis, and algorithmic validation across a variety of use-cases and frameworks
  • Develop prototype workloads, tools, and automation pipelines to accelerate performance tuning and validation workflows
  • Conduct performance and accuracy evaluation of AI models on competition HW to detect and plug the gaps