Graphics/AI Performance Engineer - Staff / Senior

Senior Engineer: 6 - 12 years Staff Engineer: 8 - 15 years About the Opportunity We are hiring on behalf of our client, a leading global semiconductor and technology company, for a Graphics/AI Performance Engineer to work on next-generation GPU architectures and AI workloads. The role focuses on performance analysis, power optimization, workload characterization, and architectural improvements for graphics and AI applications.

Key Responsibilities

Develop and maintain GPU power and performance models for graphics and AI workloads. Analyze GPU architecture and microarchitecture to identify performance and power optimization opportunities. Evaluate AI workload behavior and correlate performance metrics with power consumption. Identify system bottlenecks and propose hardware, software, or algorithmic optimizations.

Collaborate with architecture, hardware, software, compiler, driver, and ML teams. Perform power-performance tradeoff studies and provide data-driven recommendations. Contribute to future GPU architecture and roadmap planning. Required Qualifications Bachelor's degree with 6+ years, Master's degree with 5+ years, or PhD with 4+ years of relevant experience.

Strong understanding of CPU/GPU architecture and microarchitecture. Experience in performance analysis, workload characterization, and system optimization. Experience with low-power ASIC design techniques. Proficiency in C/C++ and Python. Experience with AI/ML frameworks such as TensorFlow or PyTorch. Knowledge of GPU computing technologies such as CUDA, OpenCL, Vulkan, OpenGL, or DirectX.

Experience with power analysis tools such as PrimeTime PX, Power Artist, or similar tools. Understanding of GPU drivers, compilers, and runtime software is a plus. Exposure to Verilog/SystemVerilog is desirable.

Preferred Skills

Graphics performance analysis. AI workload optimization. GPU architecture and performance modelling. Low-power design methodologies. Hardware-software co-optimization. Performance benchmarking and profiling.