GPU/CPU Integration Architect

CPU Architect Senior (L5 equivalent) 5-8 years

✓ Which company are you aiming at?

To generate precise rejection-prevention feedback and application strategy, tell us where you're targeting.

Apple

NVIDIA

Qualcomm

Intel

AMD

Arm

2 Send your application

You need an account so we know where to send the feedback.

Create an account Sign in

The job description

Tech stack. Heterogeneous system architecture, cache coherency, shared virtual memory, Arm and RISC-V ISAs, SystemVerilog, performance modeling, interconnect protocols, unified memory

About the role

You will architect the integration between CPU and GPU at a semiconductor company building heterogeneous processors for AI, graphics, and high-performance computing. The boundary between CPU and GPU is where some of the hardest system problems live: memory coherence across different consistency models, unified virtual memory, fine-grained task dispatch, and quality of service between latency-sensitive CPU threads and throughput-hungry GPU kernels. You will define how these processors work as one system rather than two chips sharing a package.

What you will achieve

Define the coherency architecture between CPU and GPU domains, selecting protocols and consistency guarantees that enable efficient sharing without imposing impossible verification burdens.
Architect unified memory and shared virtual memory schemes including page migration policies, fault handling, and TLB coherence, validated through modeling and workload analysis.
Design the dispatch and synchronization mechanisms for fine-grained CPU-GPU cooperation, minimizing launch latency and enabling programming models that treat the system as unified.
Model heterogeneous workload performance across candidate integration options, quantifying the benefits of tighter coupling against its costs in complexity, power, and area.
Specify the hardware-software interfaces for heterogeneous compute including driver-visible registers, interrupt schemes, power management coordination, and debug support.
Partner with GPU architecture, CPU architecture, software, and system teams to resolve cross-domain tradeoffs with decisions grounded in application-level performance data.

What you will bring

Must-haves

5 to 8 years of experience in CPU or GPU architecture, SoC architecture, or heterogeneous system design with shipped products.
Strong understanding of both CPU memory models and GPU execution and memory models, and the challenges of bridging them.
Experience with cache coherency protocols and their extension across heterogeneous agents.
Knowledge of GPU programming models such as CUDA, ROCm, or Vulkan compute and their hardware implications.
Performance modeling skills applied to heterogeneous workloads with the discipline to validate against real applications.
Ability to write specifications spanning multiple architecture teams with different cultures and assumptions.
Systems-level thinking and the communication skills to align CPU, GPU, and software organizations behind shared decisions.

Nice-to-haves

Experience with chiplet-based heterogeneous integration or advanced packaging.
Familiarity with AI accelerator architectures and their memory system requirements.
Background in interconnect or fabric architecture for heterogeneous SoCs.