Punzilo

Logo
view question papers
Exam Attempt
Study Mode
Question Ask Gemini

Inspiration

Growing up, I benefited from teachers who could give me close, individual attention. Classes were small enough that a teacher could notice when I was struggling, explain my mistakes in detail, and adjust how they taught me. That experience shaped how I learned.

As education in Zambia became free, access improved significantly—but class sizes also grew. With higher student–teacher ratios, it became harder for teachers to give each learner the attention they need, which risks lowering the overall quality of education.

At the same time, I noticed a deeper issue: studying hard doesn’t always translate to studying effectively. Many students passively consume content, repeat the same mistakes, and receive little or no feedback explaining why an answer is wrong. Punzilo was inspired by the need to scale personalized guidance and feedback in an education system where human attention is increasingly limited.

What it does

Punzilo is an AI-powered learning platform designed to help students learn more effectively through:

Topic-based practice using past exam questions
Instant, explanatory feedback instead of simple right/wrong grading
Adaptive difficulty based on a student’s performance
Progress and mastery tracking over time

Rather than focusing only on scores, Punzilo emphasizes understanding—helping students identify misconceptions and guiding them toward improvement.

How we built it

The project started during the hackathon with the bare minimum: basic question delivery and answer checking. As the event progressed, the idea evolved rapidly into a more complete learning system.

The platform was built around:

A frontend for student interaction
A backend API for managing users, questions, and progress
An AI layer for evaluation, feedback generation, and personalization

As ideas grew, Gemini played a major role in accelerating development. It helped with:

Rapid prototyping of new features
Generating structured feedback logic
Extracting and structuring educational content
Iterating quickly on system design

What started as a simple prototype grew into a richer platform as new tools and features became feasible within the hackathon timeline.

Challenges we ran into

One of the biggest challenges was data extraction. Most past exam papers and textbooks were only available as physical copies. To make them usable:

Documents had to be scanned
Text had to be cleaned and validated
Gemini was used to extract structured data
The output was converted into JSON for system use

This process was time-consuming and error-prone, requiring multiple checks to ensure questions, answers, and topics were correctly captured.

Another challenge was ensuring AI-generated feedback remained accurate, curriculum-aligned, and easy for students to understand, rather than overly generic or misleading.

Accomplishments that we're proud of

Building a functional, AI-powered learning platform within a hackathon setting
Successfully converting physical educational materials into structured digital data
Designing a system that prioritizes understanding and feedback over raw scores
Collaborating with teachers and students to shape features around real learning needs
Leveraging Gemini to significantly speed up development and iteration

What we learned

This project taught us that effective learning systems are built at the intersection of:

Technology
Pedagogy
Real-world constraints

We learned the importance of collaboration by working closely with teachers and students to understand how they teach and learn. Technically, we gained experience in data pipelines, AI-assisted development, and building adaptive systems. Most importantly, we learned that feedback and consistency matter more than content volume.