Inspiration
Most sports apps are built around one question: who is going to win?
For Team USA on the road to LA28, we thought there was a more interesting and more inclusive question: who are fans not noticing yet?
Olympic and Paralympic stories begin long before the Games — in hometown programs, qualification events, world championships, local clubs, and emerging fan communities. But attention is often uneven. A few sports dominate the conversation, while many Paralympic, emerging, and underfollowed sports become visible only when the Games are already here.
Draft USA was inspired by the idea that fan discovery itself could become playable. Instead of predicting medals or ranking athletes, we wanted to build a fantasy-style experience that helps fans explore Team USA momentum, Olympic–Paralympic parity, and hometown connections in a responsible, engaging way.
What it does
Draft USA: The LA28 Momentum League is a Gemini-powered fan engagement app where users draft Team USA sports, Olympic and Paralympic pairings, and hometown hubs — not individual athletes.
Each sport has a transparent Momentum Score based on five fan-facing signals:
- News momentum
- World championship activity
- Qualification context
- Hometown breadth
- Olympic–Paralympic parity
As users draft sports, they unlock bonuses for balanced Olympic and Paralympic representation, hometown diversity, and underfollowed sport discovery. The score is designed as a fan-engagement signal, not a prediction of medals, athlete performance, or competition outcomes.
Gemini acts as the app’s AI Scout and Commissioner. It recommends what kind of sport a user could draft next, explains roster parity, generates scouting reports, explains leaderboard rankings, and writes Commissioner Recaps that turn the league into a fan-readable story.
The product has four main experiences:
- Draft Room — users build a roster of Team USA sports and watch their Momentum Score update.
- Gemini Scout — Gemini gives roster-aware advice using the current draft state.
- Roster Dashboard — users see charts, score breakdowns, parity analysis, and hometown hub coverage.
- League + Commissioner Recap — users compare teams, see archetype badges, and generate Gemini-written league storylines.
How we built it
We built Draft USA as a Next.js 14 + TypeScript web app using the App Router. The frontend uses Tailwind CSS, shadcn-style component primitives, lucide-react icons, and Recharts for the momentum and roster visualizations.
The app is deployed on Google Cloud Run using a Dockerfile and Next.js standalone output.
The main external API is the Gemini API, called server-side through @google/generative-ai using gemini-2.5-flash. Gemini powers:
- Scout recommendations
- Roster analysis
- Parity coaching
- Leaderboard explanations
- Commissioner Recaps
- Safety review and responsible-language checks
The app also exposes its own Next.js API routes, including:
/api/gemini/scout/api/gemini/recap/api/gemini/safety/api/draft
We intentionally keep Gemini calls server-side so the client never sees the API key. The AI layer is wrapped in a strict agent contract that requires conditional language like “could indicate,” “may suggest,” and “could help fans discover.” It also blocks medal predictions, gambling language, guaranteed performance claims, and individual athlete rankings.
The scoring system uses local JSON seed data for sports, hometown hubs, and demo league teams. The scoring formula is documented and visible in the product so users can understand how the Momentum Score works instead of treating the AI as a black box.
Challenges we ran into
One of the biggest challenges was designing an experience that felt like fantasy sports without crossing into prediction, betting, or athlete-ranking territory. We wanted the product to feel competitive and fun, but the core mechanic had to stay focused on fan discovery, not outcomes.
Another challenge was making Olympic and Paralympic parity feel central rather than cosmetic. We did not want Paralympic representation to be a filter, sidebar, or compliance note. So we made parity part of the scoring model, the roster feedback, the Gemini Scout prompts, the leaderboard archetypes, and the Commissioner Recap.
We also had to balance AI creativity with responsible constraints. Gemini needed enough context to generate useful, exciting analysis, but also clear guardrails to avoid unsafe or non-compliant language. That led us to build both a prompt-level agent contract and a post-generation safety layer.
Finally, we had to make the demo work reliably. Since judging environments can vary, we added graceful fallback behavior so the app can still demonstrate the full experience even if a Gemini API key is not configured locally.
Accomplishments that we're proud of
We are proud that Draft USA turns a serious fan-engagement problem into something intuitive and playable.
The strongest part of the project is that parity is not just a disclaimer — it is a scoring mechanic. Users are rewarded for building rosters that include both Olympic and Paralympic sports, represent multiple hometown hubs, and surface underfollowed sports.
We are also proud of how Gemini is used. It is not just generating decorative copy. It acts as a reasoning layer that understands the current roster, evaluates balance, explains tradeoffs, and turns structured data into fan-readable narratives.
Other accomplishments we are proud of include:
- A polished, mobile-friendly sports-tech interface
- Transparent scoring logic instead of black-box AI scoring
- Server-side Gemini integration with protected API keys
- Cloud Run deployment path
- Built-in responsible AI and safety checks
- Apache 2.0 open-source licensing
- A demo flow that clearly shows product value, technical depth, and responsible framing
What we learned
We learned that responsible AI design is much stronger when it is built into the product mechanics, not added at the end.
For example, it is easy to write a disclaimer saying “this does not predict medals.” It is more meaningful to design the entire app so users draft sports instead of athletes, receive parity bonuses instead of performance forecasts, and see Momentum Scores framed as discovery signals rather than outcome probabilities.
We also learned that Gemini is most useful when it receives structured context. The best outputs came when Gemini was given the roster, sport signals, parity state, and clear constraints. That made the responses feel specific to the user’s draft instead of generic.
Finally, we learned that fan engagement does not have to be built around prediction. Discovery, hometown pride, parity, and storytelling can be just as compelling — and more inclusive.
What's next for Draft USA
The next step for Draft USA is to make the experience more dynamic and more social.
Future improvements include:
- Connecting to richer public Team USA, qualification, news, and regional datasets
- Adding user-created leagues and shareable roster pages
- Expanding hometown hub analysis with interactive regional filters
- Adding weekly Gemini Commissioner Recaps for league activity
- Improving the Momentum Score with more transparent source links and freshness indicators
- Supporting more sports, more Para classifications, and deeper Olympic–Paralympic pairings
- Adding multimodal Gemini features, such as generating visual league cards or recap graphics
- Building a more advanced safety evaluation pipeline for every AI-generated response
Long term, Draft USA could become a fan engagement layer for the entire road to LA28: a way for fans to discover sports earlier, understand Team USA more deeply, and follow Olympic and Paralympic momentum without reducing athletes to predictions.
Log in or sign up for Devpost to join the conversation.