FinLearn-AI

Boosting Customer Engagement through Interactive AI Learning

FinLearn-AI is an AI-powered Finance Learning Assistant built to demonstrate how financial platforms like Deriv can transform static content into personalized, interactive learning journeys that increase user engagement and customer retention.

While Deriv offers high-quality ebooks through its Deriv Academy, the learning experience remains passive. FinLearn-AI turns that same content into:

• Targeted quizzes
• Custom study plans
• Summarized key insights
• A GPT-4o-powered finance tutor

Inspired by Amazon’s strategy of using Generative AI to foster customer loyalty, this project shows how businesses can use RAG-based systems to keep users engaged longer, learning more, and returning often.

Why This Matters

“I visited Deriv’s website with no prior knowledge of trading. Their ebooks were informative, but I didn’t stay long. If I could quiz myself, plan my learning, and talk to an AI tutor, I’d have stayed much longer.”

FinLearn-AI was born out of this insight:

• Turn passive readers into active learners
• Turn static content into personalized experiences
• Turn information into retention

Features

• Chat with PDFs: Ask questions about Deriv’s ebooks and get contextual answers
• Quiz Generator: Instantly generate multiple choice quizzes on any finance topic
• Study Plan Creator: Get a personalized multi day study roadmap
• PDF Summarizer: View structured summaries of each ebook with key concepts
• Answer Checker: Test your quiz responses with explanations
• Web Search Fallback: Uses DuckDuckGo tools if answers aren't found in the PDFs

Knowledge Base

FinLearn-AI is powered by the following official Deriv ebooks:

• Forex Trading
• Stock Market
• Commodities
• 7 Traits of Successful Traders
• Trading Financial Accumulator Options
• Cryptocurrency
• Synthetic Indices
• Top 10 Chart Patterns

The PDFs are parsed and chunked automatically using Agno, with pypdf used internally for text extraction.

Tech Stack

Component	Technology
Backend	Flask (Python)
LLM	GPT-4o (via OpenAI)
Vector DB	LanceDB
Knowledge Base	agno + pypdf
Search Fallback	DuckDuckGo Tools
Parsing	pypdf (via Agno)

How It Works: Agents and the RAG Pipeline

FinLearn-AI uses a modular, multi-agent architecture powered by the Agno framework. Each agent is responsible for a specific type of interaction (chatting, quiz generation, summarization, or study planning), and follows a Retrieval-Augmented Generation (RAG) workflow:

RAG Pipeline Overview

1. Query Input
   The user submits a prompt (e.g., "What are candlestick patterns?").

2. Vector Search (via LanceDB)
   Relevant excerpts are retrieved from Deriv’s PDFs using vector embeddings.

3. LLM Processing (via GPT-4o)
   The retrieved context is injected into a prompt and passed to GPT-4o.

4. Response Generation
   The model returns a contextualized, human readable answer, quiz, or plan.

5. Fallback Search (Optional)
   If needed, DuckDuckGo is used to fetch additional external information.

Agents in FinLearn-AI

Agent Name	Role
rag-agent	General Q&A on finance topics
quiz-agent	Creates quizzes based on topics
summary-agent	Summarizes full Deriv ebooks
study-plan-agent	Builds multi-day personalized plans

Each agent has access to the same knowledge base (PDFs + LanceDB) but is guided by a different system prompt tailored to its purpose.

This architecture makes FinLearn-AI highly extensible, i.e you can easily plug in new tools, swap vector DBs, or add user specific memory for personalization.

Setting it up

1. Clone the Repository, Setup Virtual Environment & Install Dependencies

git clone https://github.com/seeee3/FinLearn-AI.git
cd FinLearn-AI
python3 -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

2. Create .env File

OPENAI_API_KEY=your_openai_key

3. Run the App

python app.py

Then open http://localhost:8000 in your browser.

4. Find the below line in app.py and comment it out after the first run:

knowledge_base.load(upsert=True)

Example Use Cases

• "Explain the difference between forex and stocks."
• "Generate a 5-question quiz on cryptocurrency at medium difficulty."
• "Give me a 7-day study plan to learn chart patterns."
• "Summarize the ebook on commodities."

Acknowledgments

• Deriv.com – Official source of all trading ebooks
• Agno – Modular agent framework
• OpenAI – LLM APIs
• LanceDB – Fast vector search

License

This project is for educational and non-commercial use.
Please do not redistribute Deriv's PDF content without permission.