Has owned the full frontier post- training loop end to end: At least 3+ published research at top ML conferences - must be top few/lead authors, not just a contributorDeep expertise in LLM evaluation, benchmark design, and data quality as inputs to model training – not evaluation as a standalone functionResearch domain must be in LLM training or post- training, not speech, audio, vision, or robotics 6 - 10 years of experience in AI/ML research post- undergrad across academics and industryExperience designing benchmarks, evals, or datasets for LLM improvementIndustry research experience at an AI lab or AI infrastructure companyEngineering fluency (Python/PyTorch) to collaborate on eval infrastructureCan design experiments, inspect data, and ship quantitative analysis end- to- endEducationPhD in ML, CS, computer vision, statistics, or quantitative science from a top- tier program
