AI Systems, Language & Reasoning Models

Unconventional•Mountain View, CA

About The Position

Unconventional is rethinking computing to build a foundation for AI that is 1000x more efficient by exploiting semiconductor physics and mapping neural networks directly to device physics. As a Member of Technical Staff, Language & Reasoning Models, you will develop foundational language and reasoning models that leverage the dynamics of Unconventional’s silicon. The role focuses on mapping modern language-model behaviors directly onto the physics of the company’s hardware and demonstrating high-fidelity, large-scale language understanding and generation on an unconventional computing substrate. You will work at the intersection of NLP and reasoning research and hardware codesign. The position involves designing and scaling model architectures, rethinking sequence modeling to exploit continuous-time silicon dynamics, establishing training and evaluation methods, and collaborating with hardware designers, theorists, and system builders to codesign models with physical compute primitives.

Requirements

  • An MS/PhD or equivalent research or 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, such as transformers, SSMs, and text diffusion/flow models.
  • Hands-on experience with model scaling, including successfully designing and executing full-scale distributed training runs for large language or reasoning models and managing large compute clusters.
  • Fluency in modern deep learning frameworks such as PyTorch or JAX.
  • A proven track record of writing clean, scalable training code for large language models.

Nice To Haves

  • Experience with hardware-in-the-loop training, mixed-signal hardware, quantization, or physics-informed neural networks.

Responsibilities

  • Design, train, and scale next-generation language and reasoning architectures, including transformers, state space models, diffusion/flow models, and deep equilibrium models, for unconventional compute.
  • Rethink sequence modeling to exploit the continuous-time dynamics of silicon and reduce inefficient digital abstraction.
  • Establish training recipes, loss functions, and evaluation metrics for language comprehension, logical reasoning, and generation speed while maintaining the platform’s energy efficiency.
  • Collaborate with hardware designers, theorists, and system builders to codesign model architectures and physical compute primitives.

Benefits

  • Best-in-class health benefits
  • 401k matching
  • Truly unlimited PTO
  • Complimentary meals in the Palo Alto office
© 2026 Teal Labs, Inc
Privacy PolicyTerms of Service