Senior LLM Research Scientist

XYZ AI Inc., the company behind StealthGPT, builds language models for natural, effective writing. The Senior LLM Research Scientist will improve the models powering its product by directing and conducting research, evaluating experiments, and partnering with engineering to deploy validated improvements.

Responsibilities:

• Train and improve models. Own supervised fine-tuning, LoRA, preference optimization, distillation, and experiments with custom training objectives
• Build better training data. Curate human-written corpora and source-to-target pairs. Inspect training targets, implement quality filters, and prevent duplication and evaluation leakage
• Develop trustworthy evaluations. Measure naturalness, meaning and structure preservation, omissions, and unsupported additions. Build rubric-based LLM judges and calibrate them against human review
• Improve long-document rewriting. Investigate truncation, repetition, semantic drift, and loss of coherence or structure as document length increases
• Apply research to the product. Read relevant literature, identify promising mechanisms, test them against established baselines, and work with engineering to deploy validated improvements
• Document decisions clearly. Maintain concise experiment reports covering the hypothesis, predefined success criteria, results, limitations, and next decision