Inspiration

What it does

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for Untitled

Inspiration

We were inspired by a simple observation: students often know that an answer is wrong, but they don't know why it is wrong.

Traditional learning systems usually focus on scores, correct answers, and completed questions. AI tools can explain answers, but simply giving students the solution does not always help them understand the concept behind their mistake.

We wanted to build something different.

LUMINA was created around one question:

What if AI could understand a student's mistake and help fix the misconception behind it?

Our goal was to create an AI learning companion that doesn't simply solve problems for students, but helps them become better problem-solvers themselves.

What it does

LUMINA is an AI-powered personalized learning companion designed to understand how a student is learning and adapt accordingly.

Students can type a question, upload an image, or interact using voice. LUMINA analyzes the question and the student's response, identifies the underlying concept, and provides an explanation appropriate to their learning level.

Its key feature is the AI Misconception Detector. Instead of simply marking an answer as correct or incorrect, LUMINA attempts to identify the type of mistake, such as:

• Conceptual misunderstanding
• Formula selection error
• Calculation mistake
• Sign or unit error
• Misinterpretation of the question

Based on the detected misconception, LUMINA provides a targeted explanation and generates a new practice question to check whether the student has actually understood the concept.

The platform also includes:

• 📸 Image-based question solving
• 🗣️ Voice-based learning
• 🌍 Multilingual explanations
• 🎯 Adaptive quizzes
• 📊 Learning progress and concept mastery
• 🧠 Personalized learning paths
• 🎮 Challenge-based learning
• 👩‍🏫 Teacher insights

Instead of simply measuring "How many questions did you get right?", LUMINA focuses on "What do you understand, and what should you learn next?"

How we built it

We designed LUMINA as a combination of an interactive learning interface, an AI reasoning layer, and a personalized progress system.

The basic workflow is:

[ \text{Student Input} \rightarrow \text{AI Analysis} \rightarrow \text{Misconception Detection} \rightarrow \text{Personalized Explanation} \rightarrow \text{Adaptive Practice} \rightarrow \text{Progress Update} ]

The application interface was designed to keep the learning experience simple and student-friendly.

The AI layer processes questions and responses, extracts the relevant concepts, analyzes mistakes, generates explanations, and creates personalized practice questions.

Student progress is stored so that the system can build an evolving picture of concept mastery rather than treating every question as an isolated interaction.

We also designed the system with responsible AI principles in mind. LUMINA is intended to support learning rather than replace teachers, and important answers should be checked against trusted educational material.

Challenges we ran into

One of our biggest challenges was making the AI do more than simply generate answers.

It was relatively easy to create a chatbot that could explain a question. The harder problem was designing a system that could distinguish between different types of student mistakes and respond appropriately.

Another challenge was balancing personalization with simplicity. A learning system can become overwhelming if it presents too many recommendations, statistics, and features at once. We therefore focused on presenting the most useful next step to the student.

We also had to think about the limitations of generative AI, including incorrect responses, inconsistent explanations, privacy, and the possibility of students becoming overly dependent on AI.

These challenges helped us understand that building an educational AI product is not only about model capabilities. It is also about UX design, evaluation, safety, and responsible use of AI.

Accomplishments that we're proud of

We are most proud of turning the idea of an AI tutor into a more focused concept: an AI system that tries to understand learning mistakes, not just provide answers.

The Misconception Detector became the central feature of LUMINA because it changes the interaction from:

Question → Answer

into:

Question → Student Reasoning → Mistake → Misconception → Intervention → Practice → Mastery

We are also proud of designing LUMINA as an accessible learning experience that can combine text, images, voice, adaptive practice, and multilingual explanations.

Most importantly, the project helped us think about AI as a tool for empowering students rather than simply automating their work.

What we learned

Building LUMINA taught us that a good AI application begins with a real human problem, not with the AI model itself.

We learned about multimodal AI, prompt design, adaptive learning, user experience, data handling, and the importance of testing AI outputs.

We also learned that an AI system can be technically impressive but still fail to be useful if the experience is confusing or if the AI gives students answers without helping them develop understanding.

The biggest lesson was that responsible AI is part of the product, not an extra feature added at the end.

What's next for LUMINA

Our next goal is to make LUMINA more accurate, personalized, and useful across different learning environments.

We want to expand its curriculum coverage, improve misconception detection through more extensive testing, and develop stronger teacher tools.

Future versions could include:

• Personalized revision schedules
• More advanced concept maps
• Offline or low-connectivity learning support
• Better handwriting and diagram understanding
• Teacher-created learning environments
• More regional languages
• Verified educational knowledge sources
• Long-term learning analytics

Ultimately, we want LUMINA to grow from an AI study companion into a personalized learning system that helps every student understand not only the answer, but the thinking behind it.

Built With

• adaptive
• ai
• api
• artificial
• computer
• education
• gemini
• generative
• intelligence
• language
• learning
• machine
• multimodal
• natural
• ocr
• personalized
• processing
• vision
• voice