Conference Program Archive

Title: From Intent to Impact: Designing AI for Human-Centered Decision Intelligence

Abstract: In the age of AI, the real challenge isn’t to replace human decisionmaking but to empower it.  In human-centric fields like education and healthcare, decisions are rarely one-off events—they evolve, shaped by changing goals, feedback, and context. To support real-world long-term decision-making, AI must go beyond accuracy and become dynamic, interpretable, and aligned with human intent. This talk shares how to build AI that works with people, not just for them, capturing not only what might happen next, but why. Through deep learning and reinforcement learning frameworks designed for sequential, goal-driven behavior, I’ll showcase applications from adaptive learning to early critical care. These approaches aim to close the gap between human intent and meaningful, real-world outcomes.

The transformative impact of generative AI on journalism

This session will delve into the transformative impact of generative AI on journalism, led by Ben Heller, a GenAI Transformation Manager at Google Cloud. Attendees will learn about GenAI’s capabilities in enhancing content creation, production, and audience experiences, while also exploring the critical ethical considerations and potential risks, such as misinformation and bias. The discussion will also cover strategies for ensuring trustworthy and transparent journalistic practices in the AI era, emphasizing human oversight and continuous learning.

AI-Powered News Briefing for Everyone: Introducing BriefingSmart

In today’s fast-paced world, staying informed is both essential and overwhelming. BriefingSmart offers a transformative solution: an AI-powered mobile app that curates, summarizes, and reads the news to users—anytime, anywhere. Designed for individuals, professionals, and enterprises, BriefingSmart combines audio transcription, advanced summarization, real-time alerts, and generative AI to deliver tailored, actionable news briefings. At its core, BriefingSmart uses a modular AI architecture, integrating natural language processing (NLP), speech synthesis, and large language models (LLMs) to personalize news experiences across industries.

Whether you’re a machine learning engineer, NLP researcher, or product innovator, this session will show how AI can reshape the way we consume and act on information—empowering everyone to act on what inspires them. Learn more at https://briefingsmart.ai

Beam Up Your ML: Scaling Data Processing with Dataflow and the Apache Community

This presentation explores how Apache Beam and Google Cloud Dataflow empower organizations to tackle large-scale machine learning and real-time data processing challenges. We will delve into Apache Beam’s unified programming model, which simplifies the development of both batch and streaming data pipelines. Discover how Dataflow ML, with features like the RunInference transform for efficient model execution and MLTransform for streamlined data preparation, accelerates the deployment of complex ML workflows. We’ll showcase real-world examples of massive-scale processing. Finally, we’ll highlight the vibrant Apache Beam open-source community and provide guidance on how you can contribute and leverage its collective expertise to build innovative, scalable AI solutions.

Solving Real-World Challenges with AI and Optimization

From saving endangered sea turtles to optimizing school transportation, AI is no longer just theoretical—it’s in action. In this presentation, we’ll explore a broad range of real-world applications where AI and Optimization are driving meaningful impact. While we won’t dive very deeply into each case, you’ll get a glimpse of how these technologies are applied in wildlife conservation, public sector planning, aerospace, pandemic preparedness, and more.

The Excitement and Gotchas of AI Startups

Navigating the AI startup landscape is full of promise — and peril. This talk dives into some of the challenges and unexpected pitfalls founders face from both business and technical perspectives. Drawing on investor trends and product insights, Erkang will unpack what the market is watching for and where startups might stumble in product-market fit and scaling. He will also provide case studies, and examples on using AI to scale go-to-market. Meanwhile, Shawn will provide a developer’s lens, sharing real-world experiences using tools like Cursor, GitHub Copilot, Codex, and Claude Code — highlighting integration challenges and productivity trade-offs. Together, they’ll offer actionable lessons for AI entrepreneurs aiming to unlock opportunity while avoiding common traps.

AI for Software Engineering: How Far Are We?

In this talk, Dr. Bowen Xu will share his research at the intersection of artificial intelligence and software engineering, with a particular emphasis on AI for Code (AI4Code). The presentation will begin with a brief overview of the evolution of AI4Code as a research field. Drawing from his own work, Dr. Xu will then explore two major dimensions of AI4Code: functional and non-functional properties. For functional properties, he will discuss approaches to customizing AI models to improve code correctness. For non-functional properties, he will examine efforts to make AI for coding more energy-efficient, secure, and robust. The talk will conclude with an overview of ongoing projects at Dr. Xu’s research lab, Softmax, based at NC State University.

AI-Powered Multi-Agent Auto Trading System

We have built an advanced AI-powered trading system that can make smart trading decisions automatically. It’s based on a flexible framework called TradingAgents, which allows different expert “agents” to work together—like those focused on company fundamentals, market news and sentiment, price charts, trading execution, and risk control. At the heart of our system is a Signal-Generation Master Module (SGMM), which helps all these agents coordinate and generate trading signals. The system is designed for multiple users and runs in secure, independent environments, with connections to real trading platforms through APIs and web sockets. In early testing with real and simulated trades, this automated system has shown promising results and consistent profits in stock, forex, and cryptocurrency markets. We’re excited to invite partners and investors to join us in taking this platform to the next level.

Understanding AI Agents – Making AI models productive in the real world

AI Agents are combinations of LLMs, tools, and custom roles that can autonomously perform tasks and make decisions based on context and user input. Multiple agents can be managed together to cooperatively handle individual tasks that are part of a larger project to accomplish an overall goal.

By combining capabilities like tool access, multi-step reasoning, and real-time adjustments, agents can construct and complete complex workflows and intelligent solutions. In this presentation, we’ll look at what AI agents are, how they work, and the other key information you need to know about them.

Dr. Helen Chen (NCSU), Dr. Hai Li (Duke), Dr. Ramtin Zand (USC), Eric Zhang (Google)