KADAI

Inspiration The idea for KADAI was born from a personal frustration, one that I share with millions of Spanish speakers: the language barrier in tech education. As a programmer, I've always wanted to learn the latest emerging technologies, like LangChain or new libraries, but the best documentation and the most practical courses are almost always in English first. This forces us into a double learning curve: the language itself and the technology.

I was inspired by the idea of creating a tool that could break down this barrier. What if I could have a platform that not only teaches in my language but also learns for me? A tool that could take any dense, technical documentation and, as if by magic, transform it into an interactive, personalized, and even fun course. I wanted to build what I would have loved to have myself: an AI tutor that would allow me to learn any new skill in a practical and accelerated way.

What it does KADAI is a web-based programming learning platform powered by an orchestration of Artificial Intelligence agents. In essence, it's a personal "Course Creator."

When a user arrives, KADAI asks them 4 simple questions: their experience level, what language they want to learn, their ultimate goal, and most importantly, what theme they find fun (like "Medieval Adventure" or "Space Explorer").

With that information, KADAI automatically generates a complete, tailor-made course. The platform combines theoretical lessons, narrated by an AI professor with a thematic voice with elevenlabs, with an integrated development environment (IDE) where the user can write and run code in real-time. An AI code reviewer provides hints and feedback, acting as a 24/7 personal tutor.

How we built it For this hackathon, I focused on bringing the vision of KADAI to life. I designed the entire AI architecture, conceptualizing how the intelligent agents would use an advanced technique called RAG (Retrieval-Augmented Generation) to create courses from an expert knowledge base.

The frontend, which is the visible face of KADAI, was designed and prototyped entirely in Bolt.new, where the entire user experience currently runs. Due to the limited time, the connection between the frontend and the AI backend currently operates in a "mock" mode, but with all the logic and architecture ready to be implemented and turned into a functional reality.

Challenges we ran into My biggest challenge, as expected, was time. Being a one-person team, my main focus was on implementing the system's brain: the AI agents and the RAG logic. I made significant progress, and the system is already capable of generating complete, themed courses in JSON format.

However, the major hurdle was the final integration of this AI backend with the frontend. Currently, although the course generator works, it is not yet connected in real-time to the user interface. This is why the platform operates in "mock" mode, but it is the next and final step for KADAI to fully come to life.

Accomplishments that I'm proud of I am incredibly proud of having built a functional prototype of such a complex idea in such a short time. My greatest accomplishment is seeing the entire flow come to life:

A user answers the personalization survey.

The "Course Creator" agent successfully generates a coherent, themed curriculum in JSON format.

The platform renders the lesson and the interactive IDE based on that response.

Achieving this orchestration of AI agents to work and create a fluid user experience is a technical accomplishment I am very proud of.

What I learned The biggest lesson I learned is about the quality of context in AI systems. "Garbage in, garbage out" is a very real phrase. I understood that in a RAG architecture, the quality and curation of the knowledge source are much more important than the amount of data you provide.

I also learned to simplify and be realistic. I realized that creating such a large and complex system is an immense challenge for one person, and that while "vibe coding" gives you a great initial boost, it's not always enough. But the most important lesson was not to give up. Even though I'm not presenting a 100% finished product, this experience has given me the certainty that I can make it a reality in the future. For that, I am very grateful to Bolt for launching this hackathon and giving me the opportunity to start this journey.

What's next for KADAI The future for KADAI is exciting. My immediate plan is:

Launch a Closed Beta: I will use the MVP from this hackathon to invite the first users and get real feedback.

Expand the Knowledge Base: I will apply my method to add more programming languages, like JavaScript and SQL.

Evolve the Agents: I will make the code reviewer more conversational and give the voice professor more interactions.

The Long-Term Vision: My vision is to expand the course creation engine beyond programming. I want KADAI to be the platform where anyone can upload a book, a document, or a set of notes and get an interactive and gamified course on any imaginable topic.

Built With

• css
• editor
• elevenlabsapi
• eslint
• framer
• html
• javascript
• monaco
• motion
• node.js
• postcss
• react
• reactrouterdom
• supabase
• tailwindcss
• typescript
• vite