Machine Learning Researcher Coauthor - Bradbury Gr

Skills

Leadership Research Machine Learning PyTorch benchmarking

Job Specifications

Bradbury Group is urgently looking for 1–2 strong Machine Learning Engineers to collaborate with on a 7-person research paper (3 postdocs, 2 mathematicians, 2 ML) studying the structure and geometry of neural networks.

This is a hands-on engineering + co-author role: you will work directly with senior researchers and contribute to the paper itself. Contributors will receive full co-authorship*.

(*Fair and Standard contribution tracking and Authorship agreements followed for coauthor / ordering)

We need someone who can own the core ML infrastructure for the project:

Implement model evaluation pipelines (LLMs + vision models)
Write benchmarking tools (perplexity, structural metrics, custom eval loops)
Efficiently load, fetch, shard, and decompose large models layer-by-layer
Build data loaders and matrix loaders for large-scale experiments
Integrate cloud compute (Modal, Lambda, serverless runners, etc.)
Clean up and maintain a research-grade ML codebase
(Optional) Implement experimental algorithms developed by the math team
(Optional) Research math / ML yourself (we are open to full research roles)

If you enjoy building infrastructure that scientists actually use, you will thrive here.

There is also large potential to coauthor on other papers simultaneously, if research, and implementation ability and time commitment is available. We have 2 other projects running simultaneously.

This role exists specifically to support and co-author a high-value ML paper.

You will be expected to:

Contribute to experiment design
Help write and maintain the manuscript
Interpret results with the math team
Participate in paper strategy and structuring
Join discussions on extensions and generalizations

Required:

Strong PyTorch engineering ability
Experience building ML tooling, repos, loaders, and evaluation scripts
Ability to take ownership of infra without hand-holding
Interest in neural network structure, matrix geometry (high-level only), or model internals
Comfortable working with cloud compute / serverless
10 hrs/week, fully remote, highly flexible
People who care about the bigger picture, lab-wide systems, and taking leadership as we grow

Bradbury Group is a fully non-profit, volunteer-run AI research lab

(see our website: we operate under open-science principles).

Everyone here contributes voluntarily toward peer-reviewed publication and open-source releases.

We do not provide compensation or employment.

Strong contributors gain authorship and a long-term place in the lab.

About the Company

Bradbury Group is a non-profit AI research lab of around 25 people from top AI companies and University Labs (including, Oxford, Berkeley, Hugging Face etc), focusing on generating open-source publications of cutting edge and novel research, particularly in the computer vision domain. Our research spans novel approaches to model optimization, speedups, image generation and more across various domains from geospatial, medical and pure. Know more