Research Scientist

Highlights

• Run failure analyses on frontier models using real-world task distribution.
• Design datasets, evaluations and RL environments for vision-language models.
• Publish and extend RF100-VL benchmarks grounded in real deployments.

What they offer

• Equity
• Bonus
• Generous stipends for travel, productivity, tools, meals and office
• Up to 100% health insurance coverage
• Unlimited PTO with 2-week minimum
• 12 weeks parental leave

Ideal candidate

PhD-level vision researcher with high agency who wants work to ship as real data. You code your own experiments, communicate clearly, and care about moving frontier models on real-world tasks.

Requirements

• PhD or equivalent research experience in computer vision, multi-modal learning, ML or related field with published or shipped work
• Hands-on post-training experience: fine-tuning, RL, evaluation design or data curation for large models
• Strong engineering fundamentals and ability to write code for own experiments
• Clear written communication to researchers and annotators