Smart Learning Coach

A AI personalized coach which help us to do smart preparation

Comment

Inspiration

The inspiration for Smart Learning Coach came from my own experience as a student. While preparing for exams and learning technical subjects, I constantly faced the same problems: unstructured study routines, difficulty understanding complex concepts, long YouTube videos with no clear notes, and the extra effort required to create flashcards or quizzes for revision.

Although many AI tools exist, most of them solve only one part of the learning problem—summarization, explanation, or question generation—but not the entire learning workflow. I wanted to build a system that behaves more like a personal learning coach, guiding a student from planning -> learning -> revision -> assessment.

This project was also inspired by the Gemini course content, especially the idea of agents as specialized problem solvers. I wanted to apply those concepts to a real, meaningful student-focused problem.

What it does

Smart Learning Coach is an AI-powered study assistant built using Gemini 2.0 Flash, Python, and Streamlit. It uses a multi-agent architecture, where each agent focuses on a specific learning task.

The system provides:

Study Planner – Generates personalized, day-wise study plans based on subjects, exam date, and available hours.

Concept Explainer – Explains any topic at simple, medium, or advanced levels.

YouTube Video Summarizer – Extracts transcripts and converts long videos into structured summaries, notes, chapters, and tags.

Flashcard Generator – Automatically creates revision-ready flashcards from summaries.

Quiz / MCQ Generator – Produces exam-oriented MCQs with correct answers and explanations.

AI Coding Tutor – Explains code, helps debug errors, and generates code step-by-step.

PDF Export – Allows users to download summaries, flashcards, and quizzes as PDFs.

Together, these features form a complete learning pipeline instead of isolated tools.

How we built it

The project is built using a modular, agent-based design.

Architecture

Streamlit acts as the user interface and orchestration layer.

Each learning task is handled by a dedicated agent:

Planner Agent

Explainer Agent

Summarizer Agent

Flashcard Agent

Quiz Agent

Coding Agent

Each agent uses Gemini 2.0 Flash for reasoning and content generation, with carefully designed prompts and structured outputs.

Tools

YouTube Transcript API is used to fetch transcripts for video summarization.

ReportLab is used to generate downloadable PDFs.

A simple memory system stores session-level context for smoother interaction.

Design Principles

Separation of concerns (one agent = one responsibility)

Reusable and extensible code

Secure API handling using environment variables

Clear file structure and readable, commented code

This approach reflects real-world AI system design rather than a single monolithic prompt.

Challenges we ran into

Prompt design for agents: Ensuring each agent stayed within its role without overlapping responsibilities required careful iteration.

YouTube transcript availability: Not all videos have transcripts, so fallback handling had to be implemented.

Structured outputs: Generating consistent flashcards and quizzes required enforcing strict output formats.

Balancing simplicity and depth: Explanations needed to be helpful for beginners while still useful for advanced learners.

Performance considerations: Keeping responses fast while running multiple agent calls.

Each challenge helped refine both the technical design and my understanding of agent-based systems.

Accomplishments that we're proud of

Successfully designed and implemented a true multi-agent system

Meaningful integration of Gemini 2.0 Flash across multiple agents

End-to-end learning workflow in a single application

Clean, modular codebase suitable for extension and deployment

Practical usefulness for real students, not just a demo project

PDF export functionality for offline learning and revision

What we learned

Through this project, I learned:

How to design agent roles and responsibilities

Practical prompt engineering for structured outputs

How to integrate tools with LLM-based agents

Building scalable AI applications with modular architecture

Streamlit-based UI design for AI applications

Importance of documentation and code clarity

How theory from the Gemini course applies to real systems

Most importantly, I learned how agents can transform AI applications from simple chatbots into intelligent systems.

What's next for Smart Learning Coach

Planned future improvements include:

Text-to-Speech for audio-based learning

Long-term learning memory and progress tracking

Syllabus-based learning packs (NCERT / university)

Multi-language support

Interactive coding sandbox

Deployment on cloud platforms such as Cloud Run or Hugging-Face Spaces

Learning analytics and personalized recommendations

The long-term vision is to make Smart Learning Coach a truly adaptive, AI-driven personal tutor.

Built With

Share this project:

Updates