This month’s Research Café session features two thoughtful and timely graduate research projects that tackle big, real-world questions from very different but equally compelling angles. Jenny Mai explores how Asian Americans navigate the deeply personal and often difficult process of talking about mental health, while Prajwal Srinivas looks at how we can design smarter, more responsive AI systems that reason and remember more like people do. These talks highlight the impact of careful, human-centered research, whether it’s helping us better support one another or building technologies that can truly serve society.
Disclosures Strategies and Responses in Asian American Mental Health Narratives
Despite being the fastest growing racial group in the United States, Asian Americans and immigrants of Asian origin have the lowest help-seeking rates from support networks (family members, friends, romantic partners) and mental health services (healthcare providers and health organizations). The process of sharing mental health information is often the first step to acquiring help. Thus, this semi-structured interview study (N = 18) explores how Asian Americans with mental health conditions navigate initial and subsequent disclosures within their social networks. Results illustrate that multiple disclosure strategies were used, often shaped by the receiver's assessment, cultural stigma, and cultural norms. Theoretical and practical suggestions are discussed.
Jenny Mai
Jenny is a Ph.D. student studying Health Communication in the School of Communication and Information at Rutgers University. Her research examines how cultural determinants influence health decision-making processes in communicating mental and chronic illnesses and their effects on health outcomes. At Rutgers, she is a lab member of CommUnity Health Action Lab (CUHAL), focusing on how health is communicated and impacted throughout the lifespan. In the community, she is the Chinese American Mental Health Outreach Program Coordinator at the National Alliance on Mental Illness – New Jersey, helping the Chinese population get equitable and culturally appropriate mental health care.
Architecting an Agentic Mind: Proactive and Efficient Memory for Long-Term Reasoning in LLMs
Modern large language models (LLMs) are fundamentally limited by their finite context windows, which restrict their ability to maintain coherence and recall over long interactions. My research confronts this challenge by designing and evaluating a proactive, hierarchical memory system for LLMs. The proposed architecture combines short-term sliding window memory, episodic summarization, and a scalable vector-based long-term store, coordinated by an intelligent proactive retrieval agent. This system anticipates user needs, dynamically prioritizes information, and organizes memory for efficiency and minimal latency. Key objectives include developing dynamic salience scoring, benchmarking factual recall and multi-hop reasoning, and measuring qualitative factors such as coherence and user experience. The ultimate goal is to enable smaller, more accessible LLMs to achieve complex, multi-step reasoning typically restricted to large models. This project aims to advance the scalability and practical deployment of AI assistants for real-world applications in education, industry, and society.
Prajwal Srinivas
I am a graduate student in Computer Science at Rutgers University and serve as the co-founder and CTO of Beunec Technologies, where I lead the development of a real-time, AI-integrated cloud platform focused on user-centered innovation and secure, scalable systems. I also support the academic community as a teaching assistant at Rutgers. My technical background spans AI, backend development in Java and Python, and front-end technologies like React and Next.js. I am passionate about enhancing the impact of technology on real-world problems, with work centered on intelligent cloud solutions and AI applications. Through teaching and collaborative leadership, I strive to bridge academic excellence and practical innovation in computer science.