Explore our talks hosted by IEEE Computer Society - NC East Section. Each month, we feature research presentations with opportunities for recognition and awards.
IEEE Flyer: https://xdilab.com/Flyer_IEEE.pdf
Interest Form: https://xdilab.com/research-interest.html
Presenter: Yueqian Lin - Duke University
Abstract: Effective human-AI collaboration on complex reasoning tasks requires that users understand and interact with the model's process, not just receive an output. However, the monolithic text from methods like Chain-of-Thought (CoT) prevents this, as current interfaces lack real-time verbalization and robust user barge-in. We present AsyncVoice Agent, a system whose asynchronous architecture decouples a streaming LLM backend from a conversational voice frontend. This design allows narration and inference to run in parallel, empowering users to interrupt, query, and steer the model's reasoning process at any time. Objective benchmarks show this approach reduces interaction latency by more than 600x compared to monolithic baselines while ensuring high fidelity and competitive task accuracy. By enabling a two-way dialogue with a model's thought process, AsyncVoice Agent offers a new paradigm for building more effective, steerable, and trustworthy human-AI systems for high-stakes tasks.
Registration: Click here to Register
Presentation Link: Join the Session
Presentation Date and Time: Wednesday, Apr. 8th | 3:00–3:30 PM EST
📅 Add to Google CalendarPresenter: Brij Mohan, @ LPL Financial
Abstract: This presentation provides a practical, hands-on introduction to Agentic AI and Multi-Context Prompting (MCP), demonstrating how these emerging techniques can be applied to build intelligent, autonomous systems for real-world use cases. We explore the foundational concepts of Agentic AI—including multi-agent architectures, specialized agent roles, orchestration through LangChain and LangGraph, and advanced prompting strategies that enhance reasoning and coordination. Through a live demonstration of an AI-powered Financial Planner, we showcase how multiple domain-specific agents collaborate to deliver comprehensive financial guidance covering wealth management, education planning, tax optimization, retirement strategy, and estate planning.
Registration: Click here to Register
Presentation Link: Join Zoom Webinar
Presentation Data and Time: Wednesday, Dec. 3rd | 3:00–3:30 PM EST
📅 Add to Google CalendarPresenter: Ehsan Alam, ealam@aggies.ncat.edu
Abstract: Medical image segmentation is crucial for precise anatomical delineation in diagnostic and therapeutic procedures. Despite significant advancements in medical image segmentation models, achieving both high accuracy and fairness remains a challenging and underexplored area, as improvements in one metric often lead to reductions in the other. This study addresses the challenge of enhancing both accuracy and fairness in segmentation by mitigating demographic biases through supervised curriculum learning and progressive loss. We employ a manually annotated dataset from the Osteoarthritis Initiative (OAI), including both hip and knee radiographs. By applying various curriculum learning strategies and distinct progressive loss functions that shift focus from easier to more challenging examples, we aim to improve the models' accuracy and fairness. By considering demographic factors such as race and gender, we evaluate and mitigate biases in segmentation outcomes, leading to enhanced segmentation accuracy. Our findings contribute to the advancement of medical image analysis and the promotion of fair AI models for healthcare applications.
Presentation Link: Join Zoom Webinar
Presentation Data and Time: Wednesday, Oct. 1st | 3:00–3:30 PM EST
Registration: Click here to Register
📅 Add to Google CalendarPresenter: Pal Dave, pvdave@aggies.ncat.edu
Abstract: Cyberbullying has become a widespread social issue that impacts internet users, notably through the exploitation of visual images. In previous research, the detection of cyberbullying in images was explored but faced difficulties such as limited accuracy and reliance on multimodal techniques. We address this gap by proposing an improved method for identifying cyberbullying in images using a deep learning model. Specifically, we used the VGG19 architecture to evaluate a real-world dataset of 19,300 images related to cyberbullying, achieving superior performance compared to existing methods. Our analysis identifies critical contextual characteristics in cyberbullying images that distinguish them from standard offensive image material, such as violence or nudity. We show that VGG19 outperforms the multimodal classification model proposed in previous research, with a mean detection accuracy of 95%. These findings demonstrate the utility of convolutional neural networks (CNNs) in solving the particular issues given by contextual images of cyberbullying. Our research contributes to the development of improved techniques for combating cyberbullying in visual media.
Presentation Link: Join WebEx Webinar
Presentation Data and Time: Wednesday, April 23rd | 3:00–3:30 PM EST
📅 Add to Google CalendarPresenter: Steve Lincoln Chesney, slchesne@ncat.edu
Abstract: LoRaWAN's lightweight, long-range links expose IoT deployments to replay, jamming, injection, and key-compromise threats. We develop AI-driven intrusion detection across supervised, unsupervised, and deep models, as well as federated variants for edge/cloud placement. We highlight trade-offs among detection quality, model size, and on-device feasibility. Results indicate adaptive, lightweight AI is promising for scalable LoRaWAN IDS.
Presentation Link: Join WebEx Webinar
Presentation Data and Time: Wednesday, April 30th | 3:00–3:30 PM EST
📅 Add to Google Calendar