ML Research Intern expect in office) Internship approx

London (expect 5 days in office) Internship, flexible duration (approx 10 weeks)

Telluvian is a London-based, VC-backed pre-seed startup. We use mechanistic interpretability, our understanding of what's happening inside neural networks, to make AI cheaper and more secure. Learn more luvian.ai .

We're looking for an intern to work on the interpretability questions our research and product depend on. This is primarily a research role rather than an engineering one, but you'll need to be a competent enough engineer to run your own experiments. Confident Python is the bar; frontend skills aren't expected.

You'll join a team of fewer than ten people, and you'll be given a real research question rather than a manufactured one. You'll have someone senior to go to when you're stuck, but a lot of the work will still be yours to figure out.

This is a full time position with flexible working hours. We ask that you're in person in the London office a minimum of four days a week.

We recruit interns on a rolling basis, so start dates are flexible. Tell us when you're available.

• Run interpretability experiments on pre-trained models, and work out what the results actually mean.
• Build the tooling your own experiments need. This is the part we expect you to learn on the job.
• Communicate your findings clearly, to technical and non-technical people alike.
• Keep on top of published work, not easy in a field that moves faster than the literature.

• Currently studying, or recently graduated from, a computer science or a related STEM subject.
• Comfortable writing Python.
• Some exposure to machine learning through coursework, personal projects, or self-study.
• Curious, self-motivated, and happy working in a small, fast paced team.
• A drive towards understanding pre-trained models. This is the attribute we weight most heavily.
• Self-motivated. You'll be working with support available rather than instruction given.

• Experience with standard python frameworks i.e. Num Py PyTorch and Transformers.
• Prior research experience of any kind, including projects you can talk us through.
• Eagerness to start your own projects.
• Prior interpretability work, though we mean it when we say it isn't required.
• A long-standing interest in machine learning and computational research.
• Any exposure to serving or scaling ML models.

• Close proximity to frontier interpretability research, and direct access to the people building the company.
• Flexible working and genuine autonomy over how you approach the work.
• Direct mentorship from people doing interpretability research day to day.
• A front-row view of how an early-stage startup is built.
• A meaningful, paid project you can talk about and show off afterwards.
• Possible conversion to a return or full time offer.

We unfortunately don't have the capacity to provide visa sponsorship at this time.