LLM Post-Training Research Scientist (SFT/RLHF)

Scale AI; New York (NY); On-site; USD 181,000 - 226,000; Full time

Benefits offered by this job

Health, dental and vision coverage

Retirement benefits

Learning & development stipend

Generous PTO

Commuter stipend

Job summary

Scale AI seeks Research Scientists and Research Engineers in New York to advance post-training methods for LLMs, including SFT and RLHF, across text and multimodal data. You will collaborate with researchers to refine data curation and evaluation practices and contribute to the next generation of generative models.

The role emphasizes rigorous analysis of model behavior, development of bias mitigation strategies, and publication of results in leading AI venues.

Qualifications

• PhD or MSc in CS/ML or related field
• Strong background in deep learning, RL, and large-scale model fine-tuning
• Experience with post-training techniques such as RLHF
• Excellent written and verbal communication
• Published research at major ML conferences or journals
• Previous customer-facing experience

Responsibilities

• Research and develop post-training techniques to enhance LLMs in text and multimodal modalities
• Design and experiment new approaches to preference optimization
• Analyze model behavior, identify weaknesses, and propose bias mitigation and robustness improvements
• Publish research findings in top-tier AI conferences

Skills

SFT

RLHF

Reward modeling

Bias mitigation

Model evaluation

Publications

Communication

Customer facing

Education

PhD or MSc in CS/ML

Job description

Scale AI seeks Research Scientists and Research Engineers in New York to advance post-training methods for LLMs, including SFT and RLHF, across text and multimodal data. You will collaborate with researchers to refine data curation and evaluation practices and contribute to the next generation of generative models.

The role emphasizes rigorous analysis of model behavior, development of bias mitigation strategies, and publication of results in leading AI venues.