MALVERN, Pa. — Penn State Great Valley hosted its annual student research poster competition this semester, offering opportunities for students to summarize their work, practice their presentation skills and answer questions from faculty and guests. This year’s competition was held at the conclusion of Great Valley’s Industry Engagement Day, allowing students to showcase their work among guests from local companies.
Faculty members evaluated students’ posters and named three winners.
First place: 'Preparing Job Posting Data for Management Research: A Reproducible Pipeline for Job Content Analysis'
- Presented by Hetvi Soni, master of data analytics student
- Research team: Ricardo Reyna, master of software engineering alumnus; C. Michael Smith, master of software engineering alumnus and Doctor of Engineering student at Penn State World Campus; Darren M. Greway, master of artificial intelligence student; and Penn State Brandywine undergraduate students Victoria Martinez Garcia, Aditi Joshi and Haniya Raheel
- Faculty mentors: Luna Yang, assistant professor of business and economics at Penn State Brandywine, and Satish Mahadevan Srinivasan, associate professor of information science at Penn State Great Valley
This project focused on making raw job posting data more usable for management research. Job postings are valuable for studying labor demand and skills, but the data is often noisy, duplicated and difficult to compare across platforms. The team developed an open-source, transparent data pipeline that preserves raw provenance while cleaning, standardizing, validating and analyzing job posting data. Using a dataset of 18,807 postings across chemistry, biology and data-related roles in Pennsylvania, New Jersey, Delaware and Maryland, the researchers showed how this process can support more reliable research on labor markets, compensation, geography and talent competition.
“What interested me most about this project was the opportunity to use job posting data to understand which skills are actually in demand in the current market," Soni said. "Working on the research taught me how important transparent data collection and cleaning are for producing insights that are reliable and useful.”
Soni added that her team hopes to expand their analysis to support course recommendations so students can make better choices about future classes based on job market trends.
Second place: 'CodeViz: Multi-Agent System to Automate Data Science Tasks'
- Presented by Sai Sanjna Chintakunta, master of science in data analytics student
- Faculty mentors: Nathalia Nascimento, assistant professor of software engineering; Everton Guimarães, associate professor of software engineering
Data science has the power to extract important information from data and generate insights that can transform business outcomes. However, domain experts in fields such as healthcare may not have the programming skills to implement complex data science solutions. CodeViz, a multi-agent framework, powered by large language models, automates data science workflows by accepting the user's input in natural language. It then decomposes user queries into subtasks and routes each to a specialized agent for code generation and execution. It uses retrieval-augmented generation to provide more information from external data sources and enhance the task decomposition. CodeViz acts as a collaborative tool that complements, rather than replaces, the data scientist.
This research poster is part of Chintakunta’s master’s thesis.
“It was great to interact with different people and see them genuinely interested in the tool I built and listen to their perspective on how it can be applied in their line of work,” she said.
Third place: 'Eco Pulse AI – Reinforcement Learning for Optimized Urban Energy Harvesting'
- Presented by Sudheer Pandey, master of engineering management student
This project explores how urban sidewalks can be used to harvest kinetic energy from pedestrian footsteps and convert it into usable electricity. The main challenge is that footstep energy is highly irregular and intermittent, so this project proposes an AI-managed micro-grid that combines piezoelectric tiles, which can convert kinetic energy into electrical energy, with hybrid energy storage — a supercapacitor and battery — and a reinforcement learning controller. The system is designed to improve energy routing, reduce battery stress and support low-power smart city devices more reliably. In addition, the same pressure data from the tiles can be used to generate urban intelligence, such as footfall trends, congestion insights and safety signals.
Footstep energy harvesting can be viewed not just as an energy problem, but as an intelligent urban infrastructure problem. This project focuses on using reinforcement learning to decide how harvested energy should be routed between storage and load, how to smooth irregular power output and how to make the system more secure and resilient. It also highlights the broader value of the system in supporting decentralized smart city infrastructure powered by everyday human movement.
“I wanted to explore how AI can be applied to sustainability and smart city infrastructure in a practical way,” Pandey said. “What fascinated me most was the idea that something as ordinary as human footsteps could become both a source of energy and a source of useful urban data. Working on this project helped me better understand how engineering, data and intelligent decision-making can come together to solve real-world problems. It also strengthened my interest in building innovative, technology-driven solutions that can create meaningful impact in the future.”