AI Financial Research Agent

Inspiration

Retail investors struggle to analyze long financial reports (100+ pages). Earnings call transcripts are complex and technical. News and social media create emotional bias (panic / hype). Institutional investors use advanced research tools, but retail investors don’t. We wanted to democratize AI-powered financial research.

What it does

NovaAnalyst – AI Financial Research Agent

Reads annual report PDFs Analyzes earnings call trascripts Detects financial red flags (debt spikes, governance risks) Generates Financial Health Score (0–100) Runs Bull vs Bear AI Debate mode Simulates “What-if” scenarios Answers investor questions with structured reasoning Provides confidence scores for transparency

How we built it

Frontend React dashboard File upload system Report generation interfaceQ&A interaction panel Simulation & Debate trigger buttons

Backend FastAPI server PDF text extraction AI workflow orchestration API integration with Amazon Nova

Database PostgreSQL Stores document chunks Stores embeddings metadata Saves reports and Q&A logs

AI Layer Used Amazon Web Services (AWS) Powered by Amazon Nova 2 Lite for reasoning Used Nova Multimodal Embeddings for vector search Implemented RAG (Retrieval-Augmented Generation) architecture

RAG Flow: Document chunking Embedding generation Vector retrieval Structured reasoning synthesis

Challenges we ran into

Extracting clean text from complex PDFs Reducing hallucinations in financial analysis Designing a reliable Financial Health Score logic Handling large document context efficiently Balancing Bull vs Bear AI arguments fairly

Accomplishments that we're proud of

Built a fully functional Agentic AI system Implemented multi-step reasoning instead of simple chatbot replies Created structured financial scoring system Integrated debate mode (unique feature) Designed clean investor-focused dashboard

What we learned

How Agentic AI differs from normal LLM chat systems Practical implementation of RAG architecture Vector embeddings & semantic search Designing AI workflows using AWS Importance of explainability in financial AI systems

What's next for AI Financial Research Agent

Real-time stock market data integration News + social sentiment analysis Portfolio risk analyzer Comparative company benchmarking ML-based predictive risk scoring Multi-language support Mobile app version

Built With

• amazon
• amazon-bedrock
• amazon-nova-2-lite
• amazon-web-services
• ec2
• fastapi
• javascript
• nova-multimodal-embeddings
• postgresql
• python
• rag-architecture
• react
• rest-apis