My Olympian

Inspiration

Team USA has never had one body type, one hometown, one path, or one definition of excellence. But most sports fan experiences still make people feel like spectators looking at elite athletes from a distance. We wanted to build the opposite: a safe, inclusive digital mirror that helps a fan see where their own build could sit inside the broader Team USA story.

My Olympian was inspired by the idea that Olympic and Paralympic history should be explored together, not as separate experiences and not with Paralympic sport treated as an afterthought. The project focuses on body-type alignment, sport-family patterns, and classification context rather than medals, rankings, predictions, or comparisons to named athletes. The goal is to make Team USA feel more understandable and more personal while respecting the boundaries around NIL, privacy, and performance claims.

The core question became: what if a fan could enter a few simple biometrics and receive a data-driven, Gemini-generated story that says, “here are Team USA archetypes your profile could align with,” while giving Paralympic classifications the same analytical weight as Olympic events?

What it does

My Olympian is a Gemini-powered fan engagement app that maps a user’s height, weight, age band, and movement preference to illustrative Team USA athlete archetypes. The result is not a prediction and not a prescription. It is a conditional, fan-facing interpretation of how a person’s profile could align with historical Team USA sport families across roughly 120 years of Olympic and Paralympic history.

After a user enters their profile, My Olympian returns two closest archetype matches and lets them explore the full set of five archetypes: Reach & Rhythm, Compact Power, Aerobic Engine, Precision Control, and Explosive Pivot. Each archetype connects to both Olympic sport families and Paralympic sport families, including classification context where relevant. The app uses language such as “could align with” and “historically appears near” rather than promising performance outcomes.

The result page includes an explainable scatter visualization showing where the fan profile sits relative to the archetype centroids, a 120-year timeline that presents Olympic and Paralympic columns with equal prominence, and a Story Studio that turns the match into a short personalized throughline. Gemini generates the narrative, Gemini TTS reads it aloud, and Imagen creates an abstract silhouette card that avoids real faces, logos, or named athletes.

My Olympian also includes an analyst chat agent. Fans can ask questions about why a certain archetype matched, how a Paralympic classification works, or how sport families changed across eras. The chat agent is constrained to maintain equal Olympic and Paralympic depth, avoid real athlete comparisons, and refuse training, medical, or performance-prediction requests.

How we built it

We built My Olympian as a containerized Next.js application with React, TypeScript, and Tailwind. The experience is structured around a simple flow: collect four user inputs, calculate archetype proximity, render the visual explanation, then use Gemini to generate the fan-facing narrative and conversational analysis.

The matching engine uses a transparent archetype dataset rather than a black-box model. Each archetype has height and weight centroids, movement preferences, Olympic sport-family references, Paralympic sport-family references, classification notes, era-by-era timeline entries, and safety copy. The user profile is normalized and scored against those centroids, then the app exposes both the top matches and the remaining archetypes so the experience feels exploratory instead of deterministic.

Gemini powers the project in four ways. Gemini 2.5 Pro generates the personalized 120-year narrative. Gemini 2.5 Flash powers the analyst chat. Gemini 2.5 Flash Preview TTS reads the narrative aloud. Imagen generates an abstract, NIL-safe silhouette card for the selected archetype. Every Gemini call is wrapped in system instructions that enforce conditional phrasing, Olympic/Paralympic parity, no named-athlete comparisons, and no performance, training, or medical advice.

The app is deployed on Google Cloud Run from a Dockerized Next.js build. It uses the Google GenAI SDK and supports both AI Studio API-key mode and Vertex AI mode through environment variables. We also built deterministic fallbacks for narrative, audio, and image surfaces so the UI remains testable end to end even when model credentials or quotas are unavailable during development.

Challenges we ran into

The hardest design challenge was making the experience feel personal without making unsafe or unsupported claims. A fan-facing biometric tool can easily sound like it is telling someone what sport they should play or what they could achieve. We addressed that by making the entire product conditional, explanatory, and exploratory. The app frames every result as historical alignment, not athletic destiny.

A second challenge was Paralympic parity. It was not enough to add a Paralympic note at the bottom of an Olympic-centered experience. We had to redesign the timeline, archetype cards, prompts, and chat behavior so Paralympic sport families and classifications appear at the same level of structure and depth. We also had to handle pre-1960 history honestly by explaining that the Paralympic Games had not begun yet instead of inventing missing data.

A third challenge was NIL and representation safety. We wanted a visual and narrative experience that felt emotionally compelling, but we could not rely on famous athlete names, faces, logos, or direct comparisons. The solution was to use abstract archetypes, sport families, classification context, geometric silhouette imagery, and generalized historical patterns.

The technical challenge was coordinating multiple generative surfaces inside one fast demo flow. Narrative generation, chat, TTS, image generation, Cloud Run deployment, and local fallback behavior all had to work together without making the app feel fragile. We spent a lot of time making the model calls modular, the prompts strict, and the UI resilient.

Accomplishments that we're proud of

We are proud that My Olympian presents Olympic and Paralympic context as equal parts of one Team USA story. The timeline gives both histories equal structure, the analyst chat is prompted to answer Paralympic classification questions with the same depth as Olympic event questions, and the archetype cards make Paralympic sport families visible from the first result screen.

We are also proud of the product restraint. My Olympian does not name real athletes, does not imply that a body type guarantees performance, and does not turn fan curiosity into training or medical advice. It still feels personal because Gemini transforms the match into a narrative, audio experience, and visual card, but the underlying product stays bounded and responsible.

Technically, we built an end-to-end multimodal Gemini app rather than a static prototype. The project combines reasoning, conversational analysis, text generation, TTS, image generation, custom SVG visualization, a transparent matching engine, and Google Cloud Run deployment in one working fan experience.

Most importantly, we are proud that the app makes the user feel included. The central message is that Team USA excellence has many forms, many movement patterns, and many histories. A fan does not need to look like one famous athlete to find a meaningful connection to the broader Team USA journey.

What we learned

We learned that inclusive sports analytics is as much about framing as it is about data. The same biometric input can either become an unsafe prediction or a useful reflection depending on the language, interface, and constraints around it. Conditional phrasing, transparent scoring, and visible uncertainty made the experience stronger.

We also learned that Paralympic representation has to be designed into the information architecture from the beginning. If it is added late, it becomes decorative. Building the timeline, chat prompts, archetype schema, and visual hierarchy around Olympic and Paralympic parity forced us to make better product decisions.

On the technical side, we learned how to orchestrate several Gemini capabilities in a single user flow. Gemini worked best when each model surface had a specific role: Pro for the polished throughline narrative, Flash for fast analyst Q&A, TTS for accessibility and immersion, and Imagen for safe symbolic visuals. Separating those responsibilities made the app easier to reason about and easier to deploy.

We also learned that safety constraints can improve creativity. Removing real-athlete names and faces pushed us toward archetypes, abstract visuals, classification explanations, and sport-family storytelling, which ultimately made the app more original and more respectful.

What's next for My Olympian

The next step for My Olympian is a richer data layer. We would expand from illustrative archetype centroids into a deeper, auditable dataset of public Team USA sport-family patterns, eras, hometown regions, and classification references while continuing to avoid private individuals and named-athlete comparisons.

We would also add a source-backed explanation mode so fans can see why each archetype connection appears, what each classification means, and how Olympic and Paralympic sport families changed over time. That would make the app more transparent for casual fans and more useful for educators, museums, broadcasters, and Team USA storytelling partners.

Another major direction is personalization without surveillance. We want fans to compare multiple profiles, save an archetype card locally, or generate a shareable NIL-safe story without requiring accounts or storing sensitive biometric data. Accessibility improvements are also high priority, including stronger keyboard navigation, richer captions for audio, screen-reader descriptions for charts, and multilingual support.

Finally, we would extend the experience toward LA28. My Olympian could become a living Team USA fan companion: part archetype mirror, part classification explainer, part historical timeline, and part interactive analyst that helps fans understand the full Olympic and Paralympic landscape as the next Games approach.

Built With

• api
• gemini