Research Intern
Quadrillion, New York City
About the job
We're hiring a Research Intern to help us answer a question no one has answered yet: how do you train a model that is genuinely good at research?
This is hard, and largely uncharted. Very few groups have managed to teach LLMs to perform the open-ended, long-horizon reasoning that real research demands: forming a hypothesis, designing an experiment, reading the result, and deciding what to try next. We're doing it with a small team and a strong prior that research data can be used in a highly efficient and ingenious way.
As a Research Intern, you'll own a real research question rather than a ticket. Depending on your interests, that might mean turning a corpus of arXiv papers into reinforcement learning environments, designing reward signals for ambiguous, long-horizon research tasks, building tight RL loops between user data and model training, or constructing research-focused benchmarks.
What you'll do
- Post-train models for scientific and computational research tasks
- Build evaluations and benchmarks for such tasks, and use them to measure frontier models and our own
Required qualifications