Academic profile
I study language models for knowledge-intensive reasoning, with a focus on retrieval over structured and domain-specific knowledge.
I am a PhD student at Georgia Tech working at the intersection of natural language processing and structured knowledge. My research focuses on building and retrieving from formal knowledge structures, with the goal of making language models more reliable for knowledge-intensive reasoning in specialized domains such as finance.
Research interests: knowledge graphs, structured retrieval, retrieval-augmented generation, language model reasoning, financial NLP, domain-specific evaluation, and evidence-grounded AI systems.
ACL 2026 (Oral)
We build a knowledge-graph framework that extracts multi-level QA pairs from long financial credit agreements. We find that even strong long-context LLMs struggle with multi-hop reasoning, set operations, and implicit document relations.
GEM at ACL 2026
We systematically review finance-related NLP papers to map how the field uses tasks, datasets, methods, metrics, and open resources. We find that financial NLP still needs more domain-specific metrics, crisis-period testing, multilingual data, and interpretable alternatives to general-purpose LMs.