My current work investigates the post-training of LLMs, including data synthesis, preference alignment, long-context modelling, reward modelling, and scalable and reliable evaluation, with applications to multilingual and multicultural NLP, machine translation, and finance.