Stormglass AI Safety Institute

We are looking for a Research Engineer to design, build, and execute experiments related to AI safety. These experiments will typically be run in Python using cloud computing platforms such as Vast AI or AWS.

Job Description

Research Engineer
Reports to: Director of Operations
Compensation: $90,000 - $125,000 USD
Working arrangements: Remote, quarterly travel to conferences and team retreats.

In this role you will:

• Work closely with the researchers and engineers on your assigned projects to design experiments that test various aspects of the research team’s hypotheses for the research initiative;
• Write Python code to take the experiment from the design phase to a workable draft that can be run locally;
• Verify the results of the small-scale local experiments with the research team;
• Polish the rough experiment into production-ready code that can be run on cloud infrastructure; and
• Execute the experiment on large models in cloud environments.

Additionally, you will be expected to:

• Keep complete and accurate records of your experiments, including contributing to research papers and writing regular progress reports;
• Maintain a healthy codebase for each project in GitHub, with comprehensive commenting and intuitive project structure;
• Attend conferences to share results and methodology, as well as learn from other industry professionals; and
• Optimize code at every opportunity to ensure experiments remain accessible to and reproducible by those without access to large amounts of compute resources.

As a research institution, much of what we do is at the frontier of our current understanding of AI, and as a result it can be expected that the work you will do has never been done before. Having prior experience in academic research is an asset.

While you will be expected to work closely with others on your projects, there is a significant component of independent work that may often involve little direction or oversight due to the novel nature of our research. Asking for help or guidance is always encouraged, but the role will necessarily involve lots of trial and error during the initial design phase of your experiments.

Candidate Profile

The ideal candidate will have prior experience in academic research, with a strong interest in AI safety and excellent software development skills. A firm grasp of mathematics – especially as it relates to machine learning – is essential.

We require, at minimum, a Bachelor’s degree from an accredited university in mathematics, chemistry, physics, computer science, or a related field. Candidates with Master’s or Doctoral degrees are preferred for their research experience.

In short: if you can read the following two papers and follow along reasonably well, you’ll be fine.

• From Density Matrices to Phase Transitions in Deep Learning: Spectral Early Warnings and Interpretability (Hennick & Corlouer, 2026)
• Almost Bayesian: The Fractal Dynamics of Stochastic Gradient Descent (Hennick & De Baerdemacker, 2026)

Skills that will be essential for this role are:

• Python, including Pandas, Jupyter, scikit-learn, SciPy, PyTorch, and TensorFlow;
• Knowledge of machine learning and the mathematical basis thereof;
• The ability to read and interpret academic papers;
• Working independently with little or no direct supervision; and
• Excellent communication, especially around project milestones and deadlines.

About the Company

The Stormglass AI Safety Institute is a not-for-profit research organization focused on finding ways to detect, interpret, and modify the things AI models learn during training. We are based in Fredericton, the capital city of New Brunswick, Canada. 

How to Apply

All interested candidates should send an e-mail to people@stormglass.ca with their CV. Feel free to include a cover letter, links to your website or personal GitHub page, or anything else you think would help us to evaluate your application.