Junjie Hu is not new to UW: he has been an assistant professor in the Department of Bioinformatics with affiliate professor status in Computer Sciences. Now he’s moving to Computer Sciences as an assistant professor. Hu plans to explore “the behaviors of large language models, adapting them to knowledge-intensive reasoning tasks, and aligning them safely with users across diverse expertise and backgrounds.” And his work ties in closely with the Wisconsin Idea: “When we build AI systems that allow people from different linguistic, cultural, or professional backgrounds to understand one another, we’re expanding access to knowledge, education, and collaboration.”
Hometown:
Shantou, China
Educational/professional background:
- PhD/MS in Language Technologies, School of Computer Science at Carnegie Mellon University
- M.Phil in Computer Science and Engineering at the Chinese University of Hong Kong
- B.Eng in Computer Science at South China University of Technology
How did you get into your field of research?
When I started my undergrad, the idea of building AI systems that could truly communicate with humans felt like science fiction to me. Driven by my curiosity about the mechanisms behind these AI systems, I started to study machine learning from online resources like Coursera and auditing graduate-level courses during my undergrad years. After the invention of Word2Vec, I was fascinated by the idea of learning meaningful representations of human language using artificial neural networks. This led me to work on natural language processing tasks such as machine translation and question answering after moving to CMU. Along the way, I was fortunate to collaborate with many talented students and faculty in NLP and ML, which ultimately convinced me to dedicate my career to this field.
What are your areas of focus?
I have broad interests in natural language processing, a subfield of artificial intelligence that tackles a wide range of language-related tasks. My current focus is on understanding and developing foundation models to overcome critical language barriers in both human–machine and human–human communication. I am particularly fascinated by exploring the behaviors of large language models, adapting them to knowledge-intensive reasoning tasks, and aligning them safely with users across diverse expertise and backgrounds.
What main issue do you address or problem do you seek to solve in your work?
My ultimate goal is to break down language barriers so that all language users around the world can use NLP systems to communicate and collaborate more effectively. My prior work has focused on multilingual NLP, developing universal language models that can transfer knowledge, perspectives, and pragmatic skills across languages. Beyond human languages, similar language barriers exist among domain experts, across data modalities, and between natural and formal languages. My current research explores building language agents that can safely adapt to domain-specific contexts, interpret diverse modalities such as images, audio, and code, and support trustworthy real-world deployment.
Please tell us about something you’re working on in layperson’s terms, so that non-computer scientists at UW-Madison and the general public can understand what you’re passionate about.
I’m passionate about breaking down language barriers so that people everywhere can communicate and collaborate more easily. Today’s translation tools can handle simple phrases, but they often miss the nuance, context, and cultural meaning that make human communication rich. My research focuses on creating AI systems that go beyond word-for-word translation to truly understand and adapt to different languages, domains, and even forms of communication like images, sound, and computer code. In short, I want to build language technology that helps people understand each other—whether they speak different human languages, come from different professional fields, or even communicate through different media.
What’s one thing you hope students who take a class with you will come away with?
My hope is that they leave with not just technical skills, but also an appreciation for the power of language that accumulates human intelligence through generations, and the responsibility of building AI technology that serves diverse communities safely.
What was your first visit to campus like?
On my first visit to UW-Madison, I was struck by the energy on campus–students biking down University Avenue, conversations happening across every corner, and the sense of curiosity everywhere. I was extremely excited to be part of this community.
What are you looking forward to doing or experiencing in Madison?
I’m looking forward to spending time on outdoor activities, walking or biking the lakeshore, and popping over to the Memorial Union to catch live music and relax. I’m also eager to explore some farmers’ market events and local restaurants. Most importantly, make new friends and share interesting conversations.
Do you feel your work relates in any way to the Wisconsin Idea? If so, please describe how.
Absolutely. My understanding of the Wisconsin Idea is about ensuring that the benefits of the university extend far beyond the campus to improve people’s lives across the state and the world. My research on breaking down language barriers is deeply connected to that vision. When we build AI systems that allow people from different linguistic, cultural, or professional backgrounds to understand one another, we’re expanding access to knowledge, education, and collaboration.
Hobbies/other interests:
I enjoy hiking, walking, running, exploring local restaurants, and reading.