🧠 StudyAlpha-AI-Agent

Autonomous Study & Revision System

Your Personal AI Study Coach powered by Multi-Agent Intelligence

---

🚀 Google × Kaggle Agents Intensive – Capstone Project

🏆 Hackathon Writeup + Full System Documentation

---

🟦 Problem Statement

Students today struggle with learning consistency.
Effective learning requires:

• 🗂️ structured planning
• 📊 progress tracking
• 📝 performance evaluation
• ❗ early detection of weak topics

Most learners don’t know what to study next, how long to study, or when to revise.
Traditional AI tools simply answer questions — they don’t:

• plan
• evaluate
• remember
• adapt

💡 StudyAlpha solves this by acting as a complete, autonomous study coach that:

• plans the study roadmap
• generates quizzes
• evaluates understanding
• predicts weaknesses using ML
• adjusts future study sessions automatically

---

🤖 Why I Used Agents

Learning involves multiple interdependent tasks:

✔ planning
✔ quizzing
✔ evaluating
✔ memory retrieval
✔ weakness prediction
✔ revision scheduling

No single LLM call can manage this entire workflow.

Agents enable:

• 🔹 Modularity — each task handled by a specialized sub-agent
• 🔹 Delegation — one agent passes output to the next
• 🔹 Memory — mistakes & patterns influence future learning
• 🔹 Multi-step reasoning — structured learning cycles
• 🔹 Extensibility — new agents or tools can be added easily

📘 StudyAlpha behaves like a real human tutor — thoughtful, adaptive, and aware of past performance.

---

🌐 Deployment Platform: Streamlit Cloud Entry file: app.py Auto-deploy: On every push to main 🔗 Live App: https://studyalpha-ai-agent-muwkpf3edkao3iva87sxgp.streamlit.app

🏗️ Architecture Overview

StudyAlpha is a multi-agent learning ecosystem orchestrated by a central StudyOrchestrator.

---

1️⃣ StudyOrchestrator (Master Controller)

Controls the entire pipeline:

• planning
• quiz creation
• evaluation
• tracking
• revision generation

---

2️⃣ Planner Agent

Builds personalized study plans using:

• topics
• priority
• difficulty
• available hours
• duration

Outputs a multi-day optimized schedule.

---

3️⃣ Quiz Agent

• Generates 3 conceptual questions
• Uses a TF-IDF RAG Memory
• Pulls context from past mistakes & logs

---

4️⃣ Revision Agent

Generates targeted revision tasks based on:

• past mistakes
• predicted weaknesses
• forgotten topics

Optimized for spaced repetition.

---

5️⃣ Tracker Agent

Evaluates quiz answers and logs:

• accuracy
• score
• correctness patterns

Feeds data into the ML model.

---

6️⃣ Weakness Predictor (ML Model)

A GradientBoostingClassifier predicts:

• weakness probability
• likelihood of future mistakes
• topics needing urgent focus

---

7️⃣ Memory Bank (RAG System)

Stores and retrieves:

• quizzes
• explanations
• errors
• performance logs
• improvement history

Powered by TF-IDF + Cosine Similarity.

---

8️⃣ Streamlit UI (Optional)

A simple interface to:

• generate study plans
• take quizzes
• see analytics
• visualize performance

---

🔄 Architecture Summary

Plan → Quiz → Evaluate → Predict Weakness → Revise → Store Memory → Adapt Next

---

🖼️ Architecture Diagram

---

🎬 Demo and Simulation

The notebook demonstrates a full 7-day learning simulation:

• study plans
• quiz generation
• evaluation
• ML-based weakness prediction
• dynamic revision cycles
• memory retrieval
• performance graphs
• time allocation charts
• exported reports

---

📅 Study Plan Example

For topics:

• Arrays – Priority 2
• Graphs – Priority 1
• DP – Priority 2

The planner builds a 7-day optimized schedule.

---

🔄 2) WORKFLOW DIAGRAM

(Insert your workflow diagram image below — drag & drop into GitHub editor and replace the link)

📝 Quiz Generation

Example for topic DP:

• 3 conceptual MCQs
• Memory-aware grounding

---

📈 Evaluation & Tracking

Logs include:

• correct / incorrect
• explanations
• topic-wise accuracy

---

🔮 Weakness Prediction

ML model assigns a numerical probability:

Weakness Probability: 0.73  
→ High chance of confusion  

---

🔁 Revision Planning

Revision Agent outputs:

• targeted tasks
• recommended questions
• spaced repetition intervals

---

📊 Analytics & Visualizations

Includes:

• study-time distribution chart
• accuracy line graph
• mistake clusters
• topic dependency graph
• memory-retrieval maps
• revision heatmaps

---

🛠️ The Build

✨ Engineering Philosophy

StudyAlpha is:

• modular
• reproducible
• deterministic
• easy to test
• extendable

🧰 Technologies Used

• Python
• Scikit-learn
• Pandas / NumPy
• TF-IDF / Cosine Similarity
• Streamlit
• Joblib
• Logging & Tracing

🔧 Key Design Decisions

• Deterministic mock LLM (no API keys required)
• True multi-agent architecture
• RAG-based quiz grounding
• ML-based weakness prediction
• Structured modular repository

📦 Notebook Includes

• agent modules
• memory system
• planner/quiz/revision pipelines
• ML training + explainability
• analytics & visualizations
• reproducible tests
• synthetic student simulation

---

⏳ Future Improvements

• Gemini Integration for richer feedback
• Cloud deployment (API/Serverless)
• Long-term spaced repetition pipeline
• Daily streak tracking
• PDF/notes ingestion
• Leaderboards & gamification

---

📝 Final Note

StudyAlpha acts like a real AI tutor — doing more than answering questions.
It thinks, plans, evaluates, adapts, and remembers.

This system combines:

• multi-agent intelligence
• memory
• ML prediction
• structured planning

Perfectly aligned with the goals of the Google × Kaggle Agents Intensive Capstone.

---

👤 Author

Shubham Mahajan

AI & Web Developer
“Solve. Fail. Learn. Repeat.”

---

🔗 Project Links

• studyalpha_video_demo.ipynb
• studyAlpha_demo.ipynb
• GitHub Repository: https://github.com/ShubhamMahajan880/studyAlpha-Ai-Agent