Kerja Sepenuh Masa, Campus Singapore - AI Hardware-Software Co-design Engineer di Canaan Inc. - Maukerja

Translate AI inference workloads into measurable architecture requirements and implementable accelerator designs

Responsibilities:

• Characterize Transformer inference workloads, including attention, feed-forward layers, prefill, and decoding; identify compute, memory, bandwidth, and latency bottlenecks.
• Build reproducible performance and capacity models to compare dataflows, tiling strategies, memory hierarchies, and compute microarchitectures.
• Evaluate architectural trade-offs using simulation or analytical methods
• Mapping LLM into hardware
• Work with compiler, RTL, verification, and physical design teams to define interfaces, implementation constraints, and validation cases

Required Qualifications:

• Bachelor's degree or above in computer engineering, computer science, electrical engineering, or a related field.
• Knowledgeable in computer architecture, such as AI Accelerator, CPU, network-on-chip, memory hierarchy
• Familiarity with LLM inference, quantization, compiler mapping, is a plus
• Strong learning capability, good communication and collaboration skills
• Experience with AI accelerators, GPUs, or hardware-software co-design is a plus

Preferred Qualifications:

• Projects involving AI workloads, accelerator kernels, or compiler tools
• Exposure to performance modeling, quantization, or hardware interfaces

Growth opportunities:

• Opportunities to shape the leading-edge AI chip architecture
• Learn quantitative methods for evaluating throughput, latency, memory traffic, and numerical trade-offs