Linda Ji, a master’s student in the Applied Data Science program at the University of Chicago, spent a week this summer competing in a hackathon co-hosted by United Airlines and the University of Chicago. Her team set out to build an AI-powered dashboard that could help the airline anticipate sudden spikes in passenger demand by tracking signals from news headlines, weather forecasts, and social media.
Their efforts earned them first place at the Agents in the Sky hackathon, a week-long event that brought together airline professionals and MS in Applied Data Science students to explore how agentic artificial intelligence systems made up of multiple, communicating agents might change the way airlines operate. Agentic AI systems are capable of making autonomous decisions, taking initiative, and coordinating actions to achieve specific goals, often with minimal human intervention.
“We chose this problem because demand forecasting is both high-impact and time-sensitive,” Ji said. “And it’s something that could benefit from agentic AI systems.”
A Week to Build the Future
The hackathon paired UChicago students and alumni with United Airlines professionals in cross-functional teams. There were no strict boundaries, just one rule: build something useful.
Together, they worked virtually and in person to design AI-powered tools that might one day assist with everything from maintenance and scheduling to passenger communication and disruption response.
Some teams built systems that interpreted customer reviews to surface recurring complaints. Others created agents that recommended real-time crew scheduling changes during flight delays. Many of the projects focused on automation and insight, and how to use data to make decisions faster, more intelligently, and with more context.
Collaboration in Action
For Ishwar Ramesh, a data scientist at United Airlines, the hackathon offered a rare opportunity to test new frameworks in a fast-paced, collaborative environment.
“I saw the hackathon as the perfect opportunity to dive deeper into different MCP architectures and understand how agents communicate with each other,” Ramesh said, referring to Model Context Protocols, a new standard in generative AI.
Working alongside UChicago students, Ramesh’s team built a text-to-SQL LangGraph agent capable of querying real-time flight pricing data from DuckDB via API. He credited the students with introducing him to ReWOO, a token-optimizing model architecture he hadn’t previously encountered.
“We were meeting multiple times a day, even on weekends,” he said. “Everyone was pushing hard to deliver a working solution within the tight one-week deadline.”
Several of the resulting prototypes, he added, are now being evaluated for real-world application within United.
Next Stop: Implementation
Omar Naeem, a Generative AI Engineer at United and one of the event’s judges, said the quality and practicality of submissions exceeded expectations.
“We’re actively in the process of presenting the top seven projects to leadership at United,” he said. “This just illustrates how strong we found the ideas to be in terms of solving challenges at United and the airline industry as a whole.”
One project stood out in particular: Team 12’s Flight Attendant Maintenance Reporting Tool. The system used image recognition, Retrieval-Augmented Generation (RAG), and agent-to-agent communication to help flight attendants report damaged onboard items by simply uploading photos, automatically populating maintenance forms in the process.
“The students researched well, provided evidence for why this tool would help United, and cited real-world examples where flights were delayed due to maintenance issues onboard,” Naeem said.
A Shifting Landscape
As airlines continue to invest in generative and agentic AI, both Naeem and Ramesh see significant potential in internal applications, from revenue management and maintenance to HR and operations.
“With the introduction of Agentic AI in recent months, internal applications can now be integrated together, allowing for even more collaboration and insights between teams,” Naeem said. “In my humble opinion, Agentic AI will play a big role in the industry in the upcoming years.”
For Ji, the hackathon offered a glimpse into the real-world impact of applied data science—and reinforced where she wants to take her career.
“It confirmed that I want to keep working on applied AI problems,” she said, “where data science can directly improve decision-making.”
This article was written and posted originally to the Data Science Institute