SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Preference Model is building automated ML research engineering to advance frontier AI models. The company is tackling a critical bottleneck: the lack of high-quality RL training environments that reflect real-world complexity. The founding team includes veterans from Anthropic's data team who built infrastructure and datasets behind Claude, and the company is partnering with leading AI labs to push AI toward transformative capabilities.
As a Member of Technical Staff in Research & Post-training, you will work at the intersection of research and engineering to advance self-directed learning in large language models. Your responsibilities include training and evaluating models on proprietary RL environments to validate data quality, identify task coverage gaps, and close feedback loops between environment design and model capability. You will architect and optimize RL training infrastructure using frameworks like Verl and OpenRLHF, scaling systems to handle increasingly complex research workflows. You'll design, implement, and test training environments, evaluations, and methodologies for RL agents, and profile end-to-end training runs to maximize experiment throughput and shorten iteration cycles.
Required qualifications include hands-on experience running end-to-end LLM post-training pipelines with models at least 7B in size, proficiency in Python and PyTorch or JAX, experience with modern RL training frameworks, and demonstrated ability building and operating ML infrastructure at scale. Ideal candidates have experience evaluating model outputs and building reward signals, stay current on post-training research and can translate papers into code, hold strong opinions about RL training code structure for reproducibility, balance research exploration with engineering rigor, and possess strong systems design and communication skills. The company explicitly notes that candidates don't need a PhD or extensive publications; adaptability, exceptional communication, and collaboration skills are prioritized.
