Research Scientist

About the company

Our client is a fast-growing technology company.

The role

Raydar is recruiting for this role on behalf of our client. Lead applied research on how machine learning models are measured and improved, with a focus on the quality of the data behind them. You will join early as a founding member of a small research function and set direction with a high degree of independence, partnering closely with subject-matter specialists.

What you'll do

• Create and maintain benchmarks, test suites and dataset classification schemes that establish quality standards.
• Run independent investigations and publish findings to build the company's visibility in the research community.
• Pinpoint weaknesses in current language models and assemble datasets that address them.
• Collaborate with domain specialists and delivery colleagues to describe what excellent training and testing data looks like in different fields.
• Shape the research roadmap, proposing ideas, designing experiments and rallying colleagues around them.

What we're looking for

• About 5 or more years of research experience across academic and industry settings.
• A PhD in machine learning, computer science, statistics or a quantitative discipline from a leading program.
• A strong publication record at top machine learning conferences, in a lead or near-lead author role.
• Direct, full-cycle ownership of post-training work for large language models.
• Deep knowledge of model assessment, benchmark creation and data quality as they feed into model training.
• Prior research work in a corporate AI environment, such as a model developer or AI tooling provider.
• A research specialty in language model training or post-training, rather than speech, audio, vision or robotics.
• Skill in designing experiments, examining data and delivering quantitative analysis from start to finish.
• Curiosity about how models behave and the ability to suggest data-driven remedies for their shortcomings.
• Willingness to work on-site 5 days per week, with some flexibility to 4.

Bonus points

• A master's degree combined with an outstanding body of published work may be considered in place of a PhD.
• Comfort with Python and PyTorch for working alongside engineers on evaluation tooling.

Location and work model

• New York, NY or San Francisco, CA, United States
• Hybrid: 5 days in-office, flexible to 4 days
• Full-time

Compensation, as the posting describes it

• Equity