SkillDuel

Inspiration

We were inspired by a common problem many students and professionals face: learning new skills can feel slow, passive, and difficult to stay motivated with. Many online learning platforms rely on long videos or large blocks of content, which can make it harder to stay engaged or quickly test your understanding. At the same time, we noticed how effective short-form learning and friendly competition can be at motivating people.

Platforms like Duolingo and Brilliant showed us that breaking learning into small, interactive pieces makes it much easier to stay consistent. We wanted to extend that idea by combining AI-generated micro-lessons, instant quizzes, and peer challenges. The idea behind SkillDuel is simple: learn something quickly, test yourself, and challenge friends!

What it does

SkillDuel is a web-based AI-powered learning platform that helps users learn skills through short, interactive lessons. When a user joins the platform, they complete a short onboarding process that identifies their skill level and learning goals. Based on this input, our system generates a personalized learning path.

Each lesson is delivered as a short form AI-generated video on the chosen topic, followed by a quick quiz to test understanding. After completing the quiz, users earn XP and can optionally challenge a friend to take the same lesson and try to beat their score.

Challenges are asynchronous, so both users do not need to be online at the same time. Once both have completed the quiz, a head-to-head results screen compares their scores and awards bonus XP to the winner.

The platform also includes XP, streaks, badges, shops, and a learning path map to keep users motivated to return and continue progressing.

How we built it

SkillDuel was built as a modern full-stack web application using several tools designed for rapid development and AI integration.

The frontend was built using Next.js, which allowed us to create a modular and responsive user interface. We used Tailwind CSS to quickly build consistent UI components such as the learning path dashboard, lesson player, and challenge screens.

For authentication, database storage, and caching, we used Supabase, which provides PostgreSQL-based storage and built-in authentication.

One of the most interesting parts of the project is our AI lesson generation pipeline, which we built using LangGraph, Gemini API, D-ID, and Elevenlabs

The pipeline works roughly like this:

The user completes onboarding and selects a skill.
The system checks if a lesson plan already exists for that input.
If not, Gemini API generates a personalized lesson plan and script.
The script is used to generate a personalized video with D-ID and Elevenlabs.
All results are stored in cache so future requests are instant.

This approach allows us to combine dynamic AI generation with fast performance.

Challenges we ran into

One of the biggest challenges we faced was integrating multiple AI services within a limited hackathon timeframe. Our system involves several moving parts: Gemini for lesson generation, D-ID and Elevenlabs for video creation, LangGraph for orchestration, and Supabase for storage. Coordinating the flow of data between these services required careful planning, especially since each step depends on the output of the previous one. Another challenge was handling latency from AI generation. Generating videos and scripts can take time, which could slow down the user experience. To solve this, we implemented a caching system so that once a lesson is generated, it can be reused instantly for future users with similar inputs.

Accomplishments that we're proud of

One of the accomplishments we are most proud of is successfully building a working agentic AI pipeline that can generate personalized lessons dynamically. Instead of relying on static content, SkillDuel can create learning material on demand based on user input. We are also proud of the challenge system, which transforms learning into a social and competitive experience. Finally, we are proud of how polished the learning flow and interface turned out within a short amount of time. We were able to integrate onboarding, lesson viewing, quizzes, and peer challenges into one cohesive experience.

What we learned

Building SkillDuel taught us a lot about full-stack collaboration, AI integration, and product design. We learned how to structure an agent-based AI pipeline that connects multiple APIs together while still keeping the system modular and scalable. We also learned how small UX decisions can have a big impact on how engaging a learning platform feels. From a team perspective, we learned the importance of defining clear interfaces and data schemas so that different parts of the system could be built in parallel.

What's next for SkillDuel

If we continue developing SkillDuel, we want to expand both the AI capabilities and the social features of the platform.

Some future improvements include:

Adding real-time notifications for challenges
Implementing a global leaderboard and competitive seasons
Improving personalization using learning analytics and user progress data
Building a mobile version for even faster micro-learning sessions

Our long-term vision is to create a platform where learning feels fast, social, and motivating, helping people build new skills through short bursts of knowledge and friendly competition.