Neural Network Microarchitect

Rivian Rivian · Auto · Palo Alto, CA · Mechanical & Electrical Engineering

This role focuses on the microarchitecture of a Neural Network Engine (NNE) for AI inference in Rivian's vehicles. The Principal Neural Network Microarchitect will lead the definition of the NNE core, emphasizing datapath design, scheduling, quantization-aware execution, partitioning, and performance scalability for production neural network workloads. The role involves deep cross-functional collaboration with compiler, model, firmware, verification, and SoC teams, requiring strong hardware-software co-design expertise. Key responsibilities include architecting efficient compute datapaths, managing data movement, driving architectural tradeoffs, and building performance models. The role also involves defining requirements for correctness, observability, and debuggability, and influencing long-range NNE direction.

What you'd actually do

  1. Define and evolve the NNE core microarchitecture, including compute datapaths, instruction flow, scheduling strategy, quantization support, and execution efficiency for neural network inference workloads.
  2. Architect solutions that map effectively onto Rivian’s NNE hardware model, including the processing array, partitioning strategy, and coordination of instruction and data movement across the engine.
  3. Drive architectural tradeoffs across performance, power, area, utilization, latency, and scalability.
  4. Lead definition of mechanisms for efficient movement of activations, weights, and outputs through on-chip and off-chip memory pathways and DMA architecture.
  5. Partner closely with compiler, model, firmware, verification, and SoC teams to ensure neural network workloads are translated into efficient executable flows for the NNE.

Skills

Required

  • Deep expertise in computer architecture and hardware microarchitecture
  • Strong understanding of neural network inference hardware
  • Experience architecting or optimizing specialized compute engines
  • Strong knowledge of memory hierarchy and data movement
  • Experience working across hardware and software boundaries
  • Proven ability to evaluate architectural tradeoffs using modeling, analysis, and empirical workload characterization
  • Excellent communication skills
  • BS, MS, or PhD in Electrical Engineering, Computer Engineering, or a related field

Nice to have

  • Experience with automotive or safety-aware silicon development
  • Experience with neural network workloads such as CNNs, transformers, RNNs, or related model classes used in perception and autonomy contexts
  • Familiarity with instruction-driven accelerator architectures, partitioned compute fabrics, and high-efficiency memory orchestration
  • Experience defining architecture for systems that must balance throughput, determinism, power efficiency, and debuggability

What the JD emphasized

  • production neural network workloads
  • neural network inference workloads
  • hardware-software co-design is essential
  • production

Other signals

  • designing complex silicon blocks
  • neural network inference workloads
  • performance scalability
  • hardware-software co-design