AI-Powered Race Day Communication
AI-powered, map-based communication platform consolidating race day information into an immersive, multilingual digital hub for athletes.
The Challenge
The Boston Athletic Association manages one of the world's most recognized marathon events, attracting athletes from over 100 countries. Race day information — course maps, aid station details, medical tent locations, gear bag pickup, spectator guidance — was spread across PDFs, the website, and printed guides that were out of date within hours of race day changes. Athletes needed real-time, location-aware information in their own language, and the BAA's communications team was overwhelmed fielding repeat questions over email and social media that could have been answered through better self-service tooling.
Our Solution
We built a progressive web app centered around an interactive race day map that shows athletes their precise position on the course, the nearest aid stations, medical tents, gear transport buses, and family meeting zones as they approach them. An AI assistant powered by a fine-tuned language model handles athlete questions in natural language across 12 languages, pulling answers from a structured knowledge base covering course logistics, bib pickup, qualifying standards, and weather policies. Dynamic alerts pushed through the app update athletes when aid station status changes, weather conditions affect the course, or schedule adjustments occur.
Our Approach
Interactive Course Map
Built an interactive map using Mapbox GL with a custom tile layer showing the 26.2-mile course with animated waypoints for each mile marker, aid station, medical tent, and spectator zone. Athletes' GPS position is shown relative to nearest course facilities. The map renders offline after initial load for use in areas with cell congestion near the start village.
Multilingual AI Assistant
Fine-tuned a language model on the BAA's official race documentation, FAQ history, and logistics guides. Built a conversational interface that handles questions in 12 languages, routing to the underlying model with the detected language as context. The assistant falls back to the communications team for edge-case queries outside its knowledge scope, with those conversations flagged for content review and future training.
Dynamic Alert System
Built a push notification and in-app alert system that lets the BAA communications team broadcast updates to all registered athletes or to targeted segments (e.g., athletes past mile 18, wheelchair division participants). Alerts appear as banners on the map view and in a notification log, timestamped and archived for race day reference.
PWA & Offline Capability
Built as a Progressive Web App with a service worker caching strategy that pre-caches the course map tiles, athlete's bib-linked personal schedule, and the AI assistant's core knowledge base on first load. Athletes get full offline access to course logistics information even in the cell-congested areas around Boylston Street near the finish.
"Athletes from Japan, Germany, and Brazil were getting answers in their own language within seconds of asking. The map on race day was something our communications team was genuinely proud of."
Let's Build Together
Have a similar challenge? We'd love to hear about your project.