Promoting pedestrian safety.
That was Quinn Arnold ’26’s goal when crafting his predictive navigation app, SafeWalk AI. The Applied Mathematics and Statistics major used public crime data and machine learning to create a system that ensures pedestrians know their safest route when walking through a city.
“SafeWalk AI not only serves as a proof-of-concept for predictive pedestrian routing but a template for data-driven decision systems that balance safety, efficiency, and accessibility in complex environments,” says Arnold, noting that the app can assist vulnerable populations, including tourists, night shift workers, the elderly, and others.
Trained on historical crime patterns, lighting conditions, and urban features, SafeWalk AI models safety across space and time. Arnold applied a severity score (violent, nonviolent, sexual) to individual crimes and used a geospatial indexing system to create hexagons that he mapped to a grid of Earth. Depending on a city’s crime density, he would adjust the hexagons’ size; for instance, if there were more crimes in an area, the hexagons became smaller with more detailed predictions. He then employed a math model to help estimate the number of incidents that might be seen in this area going forward.
RELATED ARTICLE: Bryant accelerates AI literacy, application for companies
Currently, SafeWalk AI has been rolled out in New York City, Chicago, Philadelphia, Boston, and Providence — with San Francisco pending. On the app, users enter their starting point and destination and receive two route options: the ‘shortest’ route and the ‘safe’ route. Arnold explains that both list how long it will take to walk and as well as distance, with the safe route highlighting the predicted risk reduction. If there is no sign of significant risk reduction, the app will notify the individual. The user then selects their preferred route and can either follow the map or use turn-by-turn navigation.
Arnold’s inspiration for the app came from Bryant’s “AI Innovation Studio,” course where the final project was to use AI for social good by developing an AI-enabled product that solved a meaningful problem. With no previous experience developing an app or working with geospatial engineering, Arnold threw himself into what quickly became a passion project.
“This project taught me that building AI isn’t about accuracy. It’s about responsibility,” reflects Arnold. “I learned how bias hides in data, how validation can lie, and how infrastructure can fail when people need it most.”
In addition to his work on SafeWalk AI, Arnold recently used his machine learning skills at his Rhode Island Novelty summer internship. There, he designed and deployed systems that blended machine learning, marketing analysis, and behavioral insights. He also focused on building a high-accuracy image classification model, a Python-based sales forecasting tool, and a vector-based query system that generated Python code using OpenAI and Gemini APIs. Furthermore, he rebuilt ecommerce tracking with GA4, GTM, and JavaScript, analyzed customer behavior with SQL and Excel, and boosted email marketing performance through UTM-based tracking and segmentation.
Graduating in December of 2026, Arnold recently accepted a job offer from Mapfre Insurance’s advanced analytics team where he will focus on machine learning. Until then, Arnold plans to focus on deepening the science behind SafeWalk AI by refining how he measures crime reduction along routes, testing different models and training datasets (including graph neural networks), and then expanding to new cities where the system proves reliable on truly held-out data. Future plans include exploring live pedestrian density and foot-traffic estimates, selectively integrating live and crowd-sourced safety signals, investigating bias from over-policing in historical data, and experimenting with reinforcement-learning and graph-based methods so routes can adapt as conditions change while still passing strict validation checks.
RELATED ARTICLE: AI needs training to be ethical as well as accurate, Bryant expert says
Due to the amount of data it requires and the resources to upkeep servers, Arnold’s end goal is to hand SafeWalk AI off to an organization that could implement it on a wider scope. He sees the app extending its reach to the city planning and public safety realms where it could be used for informing investments in spaces where crime and foot traffic overlap.
Seeing his app up and running, Arnold shares that the experience has been extremely rewarding.
“It feels surreal to have achieved this,” says Arnold, who's also a member of Bryant’s Honors Program. “It’s a reflection of what technology can mean when we use it with empathy.”