Remote-First Senior AI Inference Hardware Architect

Positron AI; United States; On-site; USD 200,000 - 300,000; Full time

Benefits offered by this job

Health, dental, vision insurance

Life and disability coverage

Unlimited paid time off

Remote-first culture

Equity

Job summary

Positron AI is seeking a Principal ASIC Architect to define the future of AI inference hardware in the United States. You will collaborate across architecture, software, ML, systems, and silicon implementation to design next-generation accelerators for frontier models.

The role requires breadth across theory and practice, including performance modeling, architectural exploration, and close alignment with workload needs.

Qualifications

• 15+ years of experience in ASIC architecture.
• Deep expertise in AI accelerator design and high-performance SoCs.
• Hands-on experience building performance models for memory systems and interconnects.
• Strong hardware/software co-design ability.
• Excellent written and verbal communication skills.

Responsibilities

• Own architectural exploration across transformer models, MoE, sparse models, and long-context inference.
• Evaluate techniques including speculative decoding, continuous batching, KV-cache evolution, sub-quadratic and linear attention, and state-space models.
• Assess memory architectures, communication architectures, and interconnects for next-generation accelerators.
• Partner with compiler, runtime, system software, firmware, ASIC engineering, and ML research teams to align hardware direction with workloads.

Skills

ASIC architecture

AI accelerator design

Memory/interconnect modeling

Hardware/Software co-design

Excellent communication

Job description

Positron AI is seeking a Principal ASIC Architect to define the future of AI inference hardware in the United States. You will collaborate across architecture, software, ML, systems, and silicon implementation to design next-generation accelerators for frontier models.

The role requires breadth across theory and practice, including performance modeling, architectural exploration, and close alignment with workload needs.