Lectra

Inspiration

L: Learning, E: Enhancement, C: Cognitive, T: Technology, R: Real-time, A: Assistance Our inspiration came from a real and recurring challenge we observed in classrooms, students struggling to keep up with lectures while taking notes, especially those registered with accessibility services. Many of them rely on volunteer note-takers, yet coverage is inconsistent and often delayed. We asked ourselves:

What if every student had an AI-powered personal note-taker , one that never misses a word, a formula, or a concept?

That idea evolved into our project: an AI-driven lecture companion that automatically records, summarizes, and personalizes the learning experience for every student.

What it does

Throughout this project, we learned the power of combining AI, accessibility, and automation. We deepened our understanding of:

Multimodal AI : integrating audio (speech-to-text) and visual (OCR/image recognition) data to create meaningful summaries.
Web system design : building synchronized platforms for both students and admins with secure authentication and data sharing.
Inclusive design principles : ensuring the platform supports diverse learning needs through text-to-speech and adaptive learning tools.

We also learned to balance technical performance with ethical concerns like privacy, consent, and secure data handling.

How we built it

Our system has two ends : Admin and Student.

Admin End (University Side):
- The classroom hardware (camera, mic, and monitor) connects to our system.
- Using the lecture timetable, the monitor automatically activates our program at the start of each class.
- Audio and board images are recorded throughout the lecture and sent to our backend.
AI Processing (Gemini Model):
- Audio is transcribed using speech-to-text.
- Images are processed using OCR to extract board content.
- Our custom model, Gemini AI, merges the two to generate clean, formatted notes.
- The model is trained on past accessibility notes, allowing it to identify important content and follow proper formatting.
Student End (Web Portal):
- Students log in to access their lecture notes instantly after class.
- They can listen to notes using text-to-speech, or generate quizzes and flashcards based on the notes.
- The AI dynamically adjusts question difficulty according to performance.
- We also integrated a voice-interactive quiz mode: a fun, game-like feature inspired by “Who Wants to Be a Millionaire?”.

Challenges we ran into

Every innovation came with its hurdles:

Data synchronization: Ensuring seamless coordination between classroom hardware, AI processing, and the web portal.
Audio and image quality: Dealing with real-world variables like background noise, low lighting, or unclear handwriting.
Model training: Teaching Gemini AI to distinguish between relevant lecture content and filler speech required careful dataset preparation.
Scalability: Optimizing our system to handle multiple classes simultaneously without lag.
Privacy & ethics: Building a solution that respects both student and professor consent in recording and processing data.

Accomplishments that we're proud of

Building a fully functional prototype that automatically captures and summarizes live lectures.
Successfully generating AI-created notes that resemble human-written accessibility documents.
Creating dynamic quizzes that adapt to student performance.
Developing an audio-interactive quiz mode for hands-free learning.
Designing a system that’s both inclusive and scalable for universities worldwide.

What we learned

We learned how to bridge AI and education meaningfully, not just as a cool tool, but as a solution to real student needs.
We gained hands-on experience in:

Multimodal AI integration (audio + visual data).
Ethical data handling and privacy safeguards.
Building adaptive learning tools that improve over time.

Most importantly, we learned that accessibility and innovation can go hand in hand, making learning equal for everyone.

What's next for Lectra

We’re excited to continue refining Lectra by:

Adding real-time note streaming, so students can follow along live.
Integrating multilingual support for international students.
Expanding the AI model’s training dataset for even more accurate summarization.
Partnering with universities to pilot Lectra in real classrooms.

Our goal is simple: