DVA.exe

Inspiration

In today’s fast-paced financial world, making informed investment decisions is overwhelming. We wanted to create a tool that empowers everyday investors with the kind of analysis institutional traders use—delivered in a simple, understandable, and actionable way. Inspired by real-world portfolio management and multi-agent decision systems, we built DVA.exe to help users make smarter decisions without needing a finance degree.

What it does

FinSight.ai is an AI-powered investment advisor that: • Analyzes stock performance with historical trend data. • Assesses risk using volatility, drawdowns, and VaR (Value at Risk). • Uses a Large Language Model (LLM) to interpret recent news headlines for sentiment. • Forecasts future stock prices. • Brings all these agents together to provide a clear BUY / SELL / HOLD decision, along with human-readable reasoning. • Visualizes everything—from stock charts to agent conversations—so users can explore how the decision was made.

How we built it

• Frontend: Built with React.js and TailwindCSS for modern UI/UX, using Recharts for dynamic financial data visualization.
• Backend: Python FastAPI that:
• Integrates yfinance for historical price data.
• Computes technical indicators like RSI, MACD, SMA.
• Calls LLMs like Mistral/OpenAI for news sentiment analysis and final decision generation.
• Coordinates multiple agents for modular decision logic.
• APIs: Uses Finnhub for real-time news, OpenAI/Mistral for LLM evaluations.

Challenges we ran into

• Parsing and syncing diverse data sources in real-time.
• Designing agent coordination to mimic human decision-making workflows.
• Handling model output formatting and edge-case errors.
• Creating a UI that balances clarity and detail for financial novices and pros alike.

Accomplishments that we're proud of

• Successfully integrated 4 different intelligent agents (trend, risk, forecasting, LLM) into a cohesive system.
• Created an interactive UI that makes financial insight visually intuitive.
• Handled natural language model outputs and transformed them into clean, trustworthy JSON responses.
• Enabled agent communication simulation to show how decisions evolve.

What we learned

• How to coordinate multi-agent architectures in real-world applications.
• How to handle noisy LLM outputs and enforce structured responses.
• Best practices in visualizing and simplifying financial data for broader audiences.
• The importance of user trust and interpretability in AI decision-making.

What's next for DVA.exe

• Add personalized investment recommendations based on user portfolio behavior.
• Integrate real-time alerts for stock changes aligned with agent insights.
• Expand agent capabilities to include ESG sentiment, earnings reports, and macroeconomic trends.
• Deploy as a browser extension or app plugin for investor platforms like Robinhood or Fidelity.