Founding AI Researcher

Description

You would be the first researcher on the team, working on models that keep improving after they are shipped.

The company builds a local AI workstation and the open source engine that runs models on it. The engine is released under Apache 2.0, and each benchmark is published on the blog along with its setup.

Responsibilities

• Research: take open questions from an idea through to a result, designing and running the experiments yourself.
• Training: post-train and fine-tune open models from start to finish on the company's own hardware.
• Evaluation: build the measurements that show whether a change actually improved a model.

Requirements

• A research background: a PhD in machine learning, completed or nearly so, or a track record that is equivalent, including first-author papers at venues such as NeurIPS, ICML, ICLR, ACL or EMNLP.
• Hands-on experience post-training models and an understanding of where compute time goes.
• Fluency in PyTorch, and the ability to read GPU kernel code when speed is the bottleneck.
• A genuine interest in running your research on hardware that people own.
• Optional: research code or released models that others use.

Conditions

Full time, hybrid, based in Milan, Italy.

Applications are reviewed on your published work and what you have built, regardless of job title.

Hiring process: a 30-minute introductory interview with a founder, a technical interview on machine learning and your research, a behavioural interview about how you work, and a technical project on a small research question. A final decision is made within a week of the project.

Visa: open only to US citizens or those with a US visa.

Experience: any level, new graduates welcome.

Benefits

• Salary of €70,000 to €120,000 gross per year, depending on experience, plus 1% to 2% equity.
• A founding role: you help set the research agenda alongside the founders and work with them directly.
• Your research goes into the open source engine and onto every the company.
• Bonus, optional: a pull request to the open source engine helps your application, including one submitted earlier. Add the link in the form and it will be reviewed before the first call. The repository is the company's site.
• If the role appeals to you but your CV does not look like a perfect match, apply anyway.