Intel seeking an AI Kernel Developer to design and optimize advanced math kernels for Intel GPUs. Collaborate with AI framework teams to accelerate AI models execution, enhance performance, scalability, and determinism, and optimize end-to-end AI workloads using C++ and Python, enabling efficient GPU utilization for next-generation AI applications.
Location: Bangalore, Karnataka, India
Job Type: Full-time
Job ID: JR0277795
Apply: Click Here
Creating and refining various AI math kernel operators that are complex for Intel GPUs to enhance compute performance and accuracy.
Working together with the AI framework and workload teams to speed-up the end-to-end AI model implementation with the help of Intel's hardware features.
Performance tuning, profiling, and numerical optimization work will be done across AI frameworks such as PyTorch, TensorFlow, or JAX.
Monitoring and upgrading current kernels with the main focus on issues of numbers stability, determinism and throughput.
Performing multi-layered systems debugging of complicated software comprising AI frameworks and GPU runtimes.
Supporting AI customers and collaborators, by offering them technical expertise, guidance, and kernel optimization assistance.
B.Tech, M.S./M.Tech, or Ph.D. in Computer Science, Electronics, or related fields.
6–12 years of experience in AI kernel or GPU programming. Proficient in Advanced C++ (C++14/17) and having intermediate Python programming skills.
Skilled in the development of machine learning kernels like GEMM, normalization, Flash Attention, or Mixture of Experts (MoE).
Good knowledge of probability theory, linear algebra, and statistics.
Knowledge of numerical formats such as float32, float16, and float8 with comprehension of trade-offs. Use of frameworks similar to PyTorch, TensorFlow, or JAX as a part of experience.
MATLAB or CUDA knowledge for GPU computation is a plus.
Practical experience with deep learning models like NLPs and LLMs is a must.
Excellent debugging, optimization, and system-level integration skills.
Superb communication and collaboration with other functions abilities.
Work with the AI framework and workload teams to hasten the end-to-end AI models via Intel’s hardware features.
Run performance tuning, profiling, and numerical optimization over AI frameworks like PyTorch, TensorFlow, or JAX Study and apply enhancements for the current kernels with concerns in numerical stability, determinism, and throughput.
Trace complex software problems in multilevel systems comprising AI frameworks and GPU runtimes.
Work together with Intel’s hardware and software engineers on the co-design of future GPU and AI architecture competencies.
Support AI customers and partners by offering technical expertise, guidance, and kernel optimization assistance.
Experience in developing and optimizing AI or math kernels with complex numeric behavior.
Exposure to random number generation kernels and low-level GPU performance tuning.
Intel Corporation is a universal company in semiconductor innovation, located in Santa Clara, California. The company was established in 1968 by Robert Noyce and Gordon Moore. Through its brilliant design and manufacturing operations, Intel has produced high-performance processors, GPUs, and AI accelerators. The company's goal is to influence the computing future by making current technology easy and accessible through AI, cloud, and edge technologies across the board of millions of devices and industries globally.