AI Hardware Architecture

The role

You will propose, design, and evaluate analog-computing hardware architectures to achieve a 1000x energy-efficiency advantage in ML inference. Work spans theoretical modeling, simulation, algorithm development, hardware/software co-design, and experimental validation.

What you will do

• Propose, design, and evaluate analog-computing hardware architectures for achieving a 1000x energy-efficiency advantage in ML inference.
• Drive invention, prototyping, and validation of the core components of a novel computing platform.
• Perform theoretical modeling and simulation.
• Develop algorithms.
• Conduct hardware/software co-design.
• Perform experimental validation in collaboration with team members.

Required qualifications

• Exceptional technical training and ability in a quantitative field (e.g., Physics, Electrical Engineering, Computer Science, or a related discipline).
• Deep knowledge of and experience in at least one of: AI hardware architecture design or integrated-circuit design.
• Exposure to and familiarity with the basic principles and engineering practice of analog and/or other unconventional physics-based computing.
• Familiarity with how modern (post-2017) AI models work and what this might imply for analog computer systems design.
• Excellent ability to communicate complex technical concepts to diverse teams across the full stack from algorithms to circuits.

Nice to have

• MS/PhD or equivalent research/project experience
• Deep understanding of the principles and engineering practice of analog and/or other unconventional physics-based computing paradigms and a proven ability to design and evaluate analog/unconventional
• Experience modeling the energy consumption of given architectural designs for AI workloads, and optimizing architectures for minimal energy consumption while maintaining ML performance
• Experience designing systems where both the AI algorithm/model and the hardware are optimized simultaneously
• Experience with designing hardware to run modern generative AI models
• Experience with detailed systems design and evaluation using at least one type of emerging non-volatile memory (e.g., RRAM), and knowledge of the tradeoffs of other non-volatile memories

Benefits

• Best-in-class health benefits
• 401k matching
• Truly unlimited PTO
• Complimentary meals in our Palo Alto office

About the company

Unconventional is rethinking computing from the ground up to build a new foundation for AI that is 1000x more efficient by exploiting the rich physics of semiconductors.