Team USA Pathfinder

Discover your Team USA Athlete Archetype — powered by Gemini AI and 120 years of Olympic and Paralympic history. Which of 42 archetypes are you?

Comment 1

Inspiration

Fan engagement with the Olympics and Paralympics has always been one-directional. You watch. You cheer. You root for your country. But you never feel like you belong in that story. USA Archetype changes that by turning 120 years of real Team USA data into a personal identity experience. The question nobody has ever asked fans is simple: which athlete are you?

What it does

USA Archetype is a fan-facing sports RPG web app powered by Google Gemini 2.5 Flash. You input your physical stats and answer questions about your mental profile and competitive strengths. Gemini analyzes your profile against real historical USA Olympic and Paralympic athlete biometric data spanning 120 years to assign you a unique Athlete Archetype identity.

Your archetype is a combination of a physical type and a mental type, producing one of 42 uniquely named combo archetypes. Each combo has its own name, custom icon, tagline, sport recommendations, and a breakdown of which Olympic and Paralympic sports your body and mind historically align with. You also get a synergy profile showing which archetypes you pair best with.

After you get your result you can share it and compare archetypes with a friend through a live comparison page that shows both cards side by side with synergy analysis.

The system is built around 7 physical archetypes crossed with 6 mental archetypes, producing 42 uniquely named identities from The Iron General to The Glider to The Puppet Master. Each has a custom-designed icon in a minimalist red, white, and blue sports RPG aesthetic.

Paralympic representation was baked into the Gemini prompt as a hard rule from day one. Every archetype result includes Paralympic sport recommendations at equal analytical depth to Olympic sports.

How we built it

The frontend is built in React using Lovable and deployed at team-usa-pathfinder.lovable.app. The backend is a Node.js and Express API deployed on Google Cloud Run. All AI analysis runs through Gemini 2.5 Flash via the Google GenAI SDK. Real 120-year USA Olympic athlete height and weight averages by sport are baked directly into the Gemini system prompt as historical context, giving Gemini genuine data to reason against rather than relying on general knowledge alone.

The backend includes server-side rate limiting, input sanitization, and structured JSON output validation to ensure Gemini returns consistent and accurate archetype assignments across all 42 combinations. The frontend includes full input validation with range checking, character limits, and a dynamic cinematic loading sequence that reflects the actual analysis happening in real time.

The compare feature generates a shareable URL that encodes the first user's archetype result so a friend can open it, see the first card, take the quiz themselves, and get a live side-by-side synergy comparison.

The visual direction was intentional from the start. The goal was a premium sports RPG meets cinematic broadcast. Dark navy background, animated starfield with a subtle US flag star cluster pattern in the upper left corner, red and white accent typography, glowing blue card borders, and 42 custom-illustrated archetype icons. Every design decision was made to make the fan feel like their archetype reveal was an event, not a form submission.

Challenges we ran into

Getting Gemini to return consistent, valid JSON across all 42 archetype combinations with exact name matching required significant prompt engineering. The combo name lookup system uses a hardcoded mapping object on the backend so the frontend always receives a verified combo name rather than relying on Gemini to generate it correctly.

Deploying a stateless Express backend to Cloud Run with proper CORS, environment variable handling, and cold start management added real infrastructure complexity. Icon hosting required moving to Cloudinary with direct URL mapping after discovering that Lovable's build system was hashing filenames and breaking local asset references.

Accomplishments that we're proud of

Building a fully working end-to-end app in under 48 hours with real Gemini AI integration, a deployed Cloud Run backend, 42 custom archetype identities, a friend comparison feature with shareable links, and full mobile support. Paralympic parity is baked into every single archetype result at equal depth to Olympic sports. The cinematic result card reveal and the compare feature make this feel like a product people would actually use and share.

What we learned

Prompt engineering for consistent, structured outputs from Gemini at scale requires more precision than expected. Managing asset hosting in a Vite-based build environment has real constraints around filename hashing that are not obvious upfront. The biggest lesson was that baking real historical data directly into the AI context window produces dramatically better and more accurate outputs than relying on general model knowledge alone.

What's next for Team USA Pathfinder

State-level athlete history showing which Team USA athletes came from your region and which sports your state has historically dominated. User accounts with archetype history so you can track how your profile changes over time. Expanded historical context per archetype. A leaderboard showing the most common archetypes across all users. Social sharing cards optimized for Instagram and Twitter. Built for the road to the LA28 Games.

Built With

Share this project:

Updates