Job description

As a Member of Technical Staff (Research) at Quadrillion Labs, you'll help train our next-generation model for research creativity. This is genuinely hard: very few people have post-trained models at the scale of several hundred billion parameters, and we're doing it with a lean team and a set of ingenious ideas — accumulating research data and training on it in an unusually data-efficient way. If you're excited about turning every arXiv paper on the internet into an RL environment, or building fast RL loops between user data and model training, this role may be for you. What you'll do - Build infrastructure for model training. You'll build our ultra-efficient infrastructure for scraping and accumulating data, turning it into RL environments, and post-training our frontier model — especially on extremely long-horizon tasks. - Develop RL techniques. No one has figured out how to train a model that's truly good at research yet. You can help us change that with mysterious RL techniques. - Construct and run evaluations and benchmarks. Build evaluations for complex, long-horizon tasks and measure both frontier models and our own against them — so we know what's working and where to push next. Required qualifications - Experience training or benchmarking large, modern models - Ability to move quickly in a zero-to-one environment Preferred qualifications - Experience training hundreds-of-billions-of-parameters frontier models #LI-EC1 About this role We work in-person 5 days a week out of our office in Midtown Manhattan. We cover generous benefits for employees, including medical, dental, and vision insurance, lunch and dinner covered, and other varied stipends. We pride ourselves on a company culture that values a diversity of thoughts and opinions. About Quadrillion Quadrillion Labs builds Qualia, an agentic research intelligence platform that does real computational research — running experiments, analyzing data, and surfacing insight for the people who do this work at the highest level. We're a team out of Stanford NLP, HRT, Citadel, Jane Street, Google Research, and Harvard Med, building for an audience that looks a lot like us. We've raised ~$18M from Conviction, Audacious, Bloomberg Beta, ex/ante, and angels including Wes McKinney, the creator of pandas. We're ~10 people in NYC, scaling to ~15 in the coming months. We think this is the Industrial Revolution moment for research, and we're building the team to define what comes next. Equal Employment Opportunity Quadrillion Labs is an equal opportunity employer. We make hiring and employment decisions without regard to race, color, religion, sex, sexual orientation, gender identity, national origin, age, disability, genetic information, veteran status, or any other characteristic protected by federal, state, or local law. We are committed to building a team where talented people can do their best work. If you need a reasonable accommodation during the application or interview process, let us know at careers@quadrillion.io.

Requirements

Must-haves

Develop RL techniques. No one has figured out how to train a model that's truly good at research yet. You can help us change that with mysterious RL techniques.
Construct and run evaluations and benchmarks. Build evaluations for complex, long-horizon tasks and measure both frontier models and our own against them — so we know what's working and where to push next.
Experience training or benchmarking large, modern models
Ability to move quickly in a zero-to-one environment
We pride ourselves on a company culture that values a diversity of thoughts and opinions.

Nice-to-haves

Experience training hundreds-of-billions-of-parameters frontier models

Application prep

Use the role’s language and application details while tailoring your materials.

Platform: ashby
Effort: quick

Benefits

Benefits detected in the listing.

Dental

Vision