Machine Learning Architect

Ayo Semiconductor•Boston, MA

2d•Onsite

About The Position

We are a venture-backed early-stage startup developing processors specialized for machine learning. The processor will provide orders of magnitude improvement in speed and power efficiency with a goal of unseating the GPU as the dominant computing platform for AI. We are seeking a deep ML practitioner to join as a founding team member - someone with hands-on experience working on or alongside a foundation model at scale, who understands what happens under the hood when splitting jobs across thousands of GPUs, and who is excited to bring that depth to novel hardware. They have a full-stack understanding of machine learning architectures, love to optimize algorithms across disciplinary boundaries, and will deploy and train models directly on our prototype chips to help us prove out what our processor can do - no prior hardware experience required.

Requirements

PhD in machine learning, representation learning, theory of computation, or a related field - or equivalent industry experience working on foundation models at scale.
Experience training models at scale - distributed training across many GPUs, working with large datasets and compute.
Has built and trained neural networks from scratch
Deep knowledge of the structure and internal operation of neural networks - including how and why they behave the way they do (e.g. interpretability or explainability work is a plus).
Excitement about applying deep AI expertise to new and novel hardware environments - you don’t need prior experience with photonics or silicon, but you want to learn.
Fluent knowledge of Python
Fluency in PyTorch (preferred), TensorFlow, JAX, or other industry-standard ML software libraries

Responsibilities

Deploy and run trained models on prototype hardware and digital twins, producing working demonstrations on our chips.
Develop and adapt algorithms to train models on novel processing environments, including our prototype hardware.
Work with hardware engineers to define and refine processor architecture based on insights learned through model training and experimentation.
Maintain a deep curiosity about what makes machine learning systems work - and bring that curiosity to bear on how they run on new hardware.
Support go-to-market strategy development