This area applies machine learning and language models to protein sequence data to predict post-translational modification (PTM) sites, such as glutarylation, malonylation, methylation, and propionylation. It combines learned sequence representations with evolutionary and structural features to improve prediction accuracy and generalization. 

This area investigates the use of large language models as automated evaluators, including rubric-based scoring, pairwise comparison, and the reliability, bias, and agreement of LLM judges with human assessment. The goal is to develop scalable and trustworthy evaluation frameworks for generative systems, applied to tasks such as code generation, question answering, and educational assessment.

This area studies small language models (SLMs) as efficient alternatives to large models, including knowledge distillation from LLMs to SLMs, quantization, and systematic analysis of reasoning failures. The aim is to enable accurate, resource-efficient, and deployable language models for educational, financial, and enterprise applications.

This area investigates retrieval-augmented and agentic architectures that combine information retrieval, multi-step reasoning, and domain adaptation. The aim is to build dependable LLM-based assistants for enterprise and digital banking applications.

This area studies methods for aligning large language models with human, cultural, and ethical values, including self-alignment and benchmark-based evaluation of model behavior. The goal is to improve the reliability, interpretability, and safety of language model outputs.