Numerai is building a self-improving hedge fund.

We run an institutional quant fund built from thousands of machine learning models. Increasingly, those models are built by AI, not people. This isn't a thesis we're pitching. It's already how Numerai gets smarter every day (see https://numer.ai).

The next step is to compound it. Numerai is forming a new team with one mandate: build AI that does quant research autonomously, all the time. Discover new features, build better risk models, improve how the fund turns thousands of signals into a portfolio.

Numerai is a small company with an unusual amount of leverage per person. No bureaucracy, no politics. Significant equity.

The Role

Your first mandate is Autoresearch: our system for having machines do quantitative research. You will make it dramatically better - better at generating hypotheses worth testing, better at running experiments correctly, better at knowing when a result is real, and better at compounding what it has already learned. This is an open research problem.

Your second mandate is everything else LLMs can do for a hedge fund. Extracting structure from unstructured text and filings at scale. Building tools our researchers actually use daily. Automating the parts of the research and engineering workflow that are currently bottlenecked on human attention.

You will build the evaluations that tell us whether any of this is working. Vibes-based assessment of an AI system that influences a portfolio is not acceptable, and designing the measurement is a large part of the science here.

You will stay ahead of the field and be responsible for translating it: when a new capability lands, you should already have a view on whether it changes what we should be building.

Requirements

• Deep, current expertise in modern machine learning, with emphasis on LLMs — training, fine-tuning, RL, inference, agents, or evaluation
• A track record of frontier work: strong publications, notable open source, or systems you built that were ahead of their time
• Demonstrated ability to take an ambitious research idea all the way to something that runs reliably and gets used
• Strong software engineering. You write code other people can build on
• Excellent written communication, and the intellectual honesty to report negative results clearly
• Genuine curiosity about markets, even if you have no finance background

Nice to have

• Experience with automated or agentic scientific discovery
• Experience building LLM evaluation infrastructure for tasks without clean ground truth
• Quantitative finance exposure