CoreAI Discovery Day: Foundational AI at UNC powering discovery across Carolina
On February 9, UNC-Chapel Hill's Department of Computer Science opened Sitterson Hall for a focused look at the work that makes AI tools possible. Faculty, students and campus leaders gathered to see how core research in algorithms, perception and robotics is moving into labs, clinics and classrooms across Carolina.
Chancellor Lee H. Roberts set the tone in brief remarks and toured five spotlight demos before the talks began. His message was simple: Carolina intends to lead in AI - and it starts with strong, interdisciplinary partnerships.
What happened
The inaugural CoreAI Discovery Day brought together computer scientists with colleagues from chemistry, geography, philosophy, pharmacy, psychiatry and radiology, plus graduate students, alumni and university leadership. Department Chair James Anderson posed the question of the day: while many talk about using tools like ChatGPT, who is building the foundations that make them work?
Distinguished Professor Mohit Bansal outlined the department's AI portfolio and growth. UNC Computer Science has expanded faculty in biometrics, deepfake detection, cyber-physical systems and society-centered AI - and is ranked eighth nationally in NLP, machine learning and computer vision by CSRankings.
From algorithms to action
Twelve lightning talks spanned three tracks: computer vision and video intelligence; machine learning foundations and language intelligence; and robotics and embodied intelligence. The focus stayed on methods that lead to measurable outcomes.
- Gedas Bertasius shared a video analysis framework that outperforms commercial models from Google and OpenAI, enabling personalized feedback for skills like basketball or piano.
- Xiaoming Liu presented advances that help people tell authentic media from AI-generated fakes - a rising need as generative tools proliferate.
- Ron Alterovitz detailed the first AI-driven medical system to autonomously steer a flexible needle through living tissue, outperforming human physicians in bronchoscopy procedures.
- Samarjit Chakraborty, in collaboration with UNC Neurosurgery, showed deep brain implants for Parkinson's disease that learn and adapt in real time.
- Other talks covered theoretical guarantees for model behavior, socially aware language models for education and mental health, AI that communities can shape and govern, and agents for precision agriculture.
Demos that made it concrete
Five hands-on demos in Sitterson's upper lobby gave attendees a direct look at the research. Visitors watched an autonomous surgical robot at work, saw 3D models reconstructed from endoscopy video, and tried systems that reason about video more effectively than leading commercial products.
Doctoral student Andrea Dunn-Beltran demonstrated the endoscopy-to-3D pipeline, while undergraduates Fletcher Stuart and Ansh Aryan showed virtual reality reconstruction research in action.
Cross-campus momentum
The Networking Lunch and AI Research Fair featured posters from RENCI, the Eshelman School of Pharmacy, and departments spanning chemistry, geography and the environment, statistics and operations research, and applied physical sciences. Projects ranged from AI-designed protein therapeutics to forest biomass mapping to adversarial attacks on financial forecasting models.
A closing panel - "Core to Application: How AI Research Enables Discovery Across Carolina," moderated by Jay Aikat of the School of Data Science and Society - dug into what it takes to move from a foundational algorithm to a deployed tool, and why guardrails matter when systems touch real lives. Senior Associate Dean Jaye Cable noted that hallway conversations sparked new collaborations on the spot.
Why this matters to researchers
- Access to core methods: Vision, language, robotics and theory work that can plug into wet labs, clinics, field research and data services.
- Pathways to deployment: From guarantees and evaluation to domain-specific interfaces, the pieces needed to go from paper to practice.
- Interdisciplinary lift: Shared infrastructure, co-advised projects and faster iteration with teams that span departments.
If you're exploring how to integrate AI into scientific workflows, see AI-focused training and tools curated for research audiences: AI for Science & Research.
What's next at Carolina
CoreAI Discovery Day aligns with a broader university strategy. Vice Provost for AI and Chief AI Officer Jeffrey Bardzell is connecting computer science research with applications across every school and college. The department plans to build on new connections with ongoing partnerships and events.
Want more details on UNC's AI initiatives? Visit unc.edu/unc-ai. The department will also host a community open house on Saturday, March 7, 2026, opening its doors to the wider Chapel Hill community and highlighting research across computer science, beyond AI alone.
Your membership also unlocks:
