AI Theory & Systems, Algorithms & Models

1 мес. назад

USASenior

neural networksdiffusion modelsenergy-based modelsneural odesdeep equilibrium models

Design and scale novel AI model architectures optimized for new computational substrates focusing on efficient AI computation.

Обязанности

As a Member of Technical Staff, AI Systems, Algorithms & Models, you will design and scale model architectures that combine novel computational primitives with conventional neural network components to achieve competitive capability across modalities. You will build scaled, end-to-end architectures for image generation, language & reasoning, video, and world models that function optimally on our novel compute substrates.
AI Algorithms & Models: Co-design and evaluate next-generation AI models, including diffusion, flow, energy-based models, Neural ODEs, and Deep Equilibrium Models. You will collaborate closely across the team to combine, modify, and implement core modeling components, including both conventional (e.g., attention, normalization, Mixture-of-Experts, FFNs) and unconventional components. You will ensure that they function optimally across our novel compute substrates.
Scaling & Capability: Establish and test scaling laws specific to our novel hardware. Develop and refine methodologies and strategies (e.g., muP)  for pushing model capability toward the frontier of conventional and newly enabled capability.
Multi-Modal Exploration: Investigate the applicability of novel architectures across modalities including image generation, video, language & reasoning, and world models. Develop general learning frameworks (e.g., flow matching) across these domains.
Cross-Functional Collaboration: Act as a translator across the organization, ensuring architectural decisions are informed by both theoretical insights from fundamental science and physical constraints from hardware. Discuss algorithmic trade-offs with theorists and convert model requirements into concrete specifications for systems and hardware teams.

Требования

Education: An MS/PhD or equivalent research/project experience in a quantitative field such as AI/Machine Learning, Computer Science, Physics, Electrical Engineering, or Applied Math.
Experience: Deep expertise in the theory, training, and empirical analysis of modern ML model architectures. Demonstrated ability to design, implement, and scale models across tasks and modalities.
Architectural Fluency: Strong understanding of both conventional architectures (Transformers, Mixture of Experts, diffusion models) and continuous-time or implicit architectures (Neural ODEs, Deep Equilibrium Models, flow-based generative models), including their computational properties and trade-offs.
Software Development: Deep experience with PyTorch or JAX, including implementation of custom model components and training loops.

Будет плюсом

Numerical Methods: Experience with numerical methods for differentiating through continuous-time systems (e.g., backpropagation through differential equations, adjoint methods, fixed-point solvers).
Unconventional Co-Design: A forward-looking perspective on co-designing model architectures for unconventional computing paradigms that map closely to the physics of underlying systems.

Условия

The Mission: Redefine computing for the next 50 years by solving the fundamental energy limitation of AI at a global scale.
The Impact: Shape the company's future as a foundational team member. Enjoy massive ownership and an outsized opportunity to drive change.
The Perks: A comprehensive package including best-in-class health benefits, 401k matching, truly unlimited PTO, and complimentary meals in our Palo Alto office.

Другое

Since 2022, AI has entered the mainstream, reshaping entire industries from education and software development to fundamental consumer behaviors. This revolution has created an unprecedented demand for computation - a demand that is now fundamentally limited by energy, not just in the datacenter, but at a global scale.
At Unconventional, our mission is to solve this. We are rethinking computing from the ground up to build a new foundation for AI that is 1000x more efficient. We're doing this by exploiting the rich physics of semiconductors, mapping neural networks directly to the device physics rather than relying on layers of inefficient abstraction.