AI Systems, Language & Reasoning Models

The role

You will drive the development of foundational language and reasoning models that leverage the dynamics of novel silicon. You will map the behaviors of modern language models directly onto the physics of the hardware.

What you will do

• Design, train, and scale next-generation language and reasoning architectures specifically tailored for unconventional compute.
• Rethink standard sequence modeling to exploit the continuous-time dynamics of silicon.
• Establish training recipes, loss functions, and evaluation metrics.
• Collaborate with hardware designers, theorists, and system builders to co-design model architecture alongside physical compute primitives.

Required qualifications

• MS/PhD or equivalent research/project experience in a quantitative field such as AI/Machine Learning, Computer Science, Physics, Electrical Engineering, or Applied Math.
• Deep, hands-on expertise in the theory, architecture, and training of modern foundation models (transformers, SSMs, text diffusion/flow, etc.).
• Hands-on, battle-tested experience dealing with model scaling.
• Successfully designed and executed full-scale, distributed training runs for large language or reasoning models, managing the complexities of massive compute clusters.
• Fluent in modern deep learning frameworks (PyTorch or JAX).
• Proven track record of writing clean, scalable training code for large language models.

Nice to have

• Experience working with hardware-in-the-loop training
• Mixed-signal hardware
• Quantization
• Physics-informed neural networks

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.