riskscope-sonar-lens

RiskScope – Know the risks, trust the reasons.

Comment

📖 About the Project

💡 Inspiration

We set out to solve a core problem in modern finance: investors often don’t know why an asset’s value changes. While tools like Bloomberg or Yahoo Finance show market data, they lack explainability and real-time context across geopolitical, regulatory, and ESG-driven events.

The turning point was discovering the Perplexity Sonar API — an AI research engine capable of answering complex, real-world questions with citations from trustworthy sources. It sparked a big idea:

"What if we built an explainable risk engine — one that tracks your portfolio or a company and tells you why it’s under threat, with real-time answers and trusted links?"

And that’s how RiskScope was born.

🧠 What We Learned

  • How to ask better financial questions and frame them for AI to extract multi-faceted risk insights
  • How to integrate and orchestrate real-time data pipelines using Perplexity Sonar for reasoning and Polygon.io for raw market data
  • How to quantify subjective risks (like sentiment, misinformation, or ESG controversies) and present them visually
  • The importance of citations and transparency when using AI in high-stakes fields like finance

🔧 How We Built It

  • Frontend: React.js + TailwindCSS, building a modular and responsive dashboard
  • Visualization: Recharts for charts and risk maps
  • Backend: FastAPI (Python) acting as the glue between Perplexity Sonar, finance APIs (like Alpha Vantage), and the frontend
  • Sonar Integration: We built a middleware that reformats user queries, fetches answers from the Perplexity Sonar API, and enriches them with:
    • Risk categories (regulatory, geopolitical, ESG, sentiment)
    • Confidence scores
    • Relevant URLs for transparency
  • Portfolio Uploader: Users can upload a CSV or connect an API key to analyze risk exposure across industries, countries, and event types

🚧 Challenges We Faced

  1. Contextualizing Financial Risk from Open-ended Queries
    Financial data is nuanced — a drop in stock price may be due to earnings misses, sanctions, or viral misinformation. Mapping Sonar responses into structured “risk types” was challenging and required custom classifiers.

  2. Real-time vs. Reliable Sources
    Some real-time news lacks credibility. We had to filter and prioritize sources in Sonar’s response pipeline using a source credibility layer.

  3. API Cost Management
    Since Perplexity Sonar performs live searches across multiple sources, optimizing queries to avoid exceeding limits or latency was critical. We implemented query caching and batching to reduce calls.

  4. Building Trust in AI Explanations
    Presenting AI-generated insights in finance can be risky if users can’t validate them. Sonar’s cited sources solved this — every explanation has a paper trail.

🚀 What Makes This Project Special

RiskScope isn’t just a financial dashboard — it’s an AI research analyst in your pocket, powered by Perplexity’s cutting-edge retrieval engine. It doesn’t give you black-box answers — it tells you the “why”, backed by real-world evidence.

In a world where trust in AI and finance is everything, we believe RiskScope is a step forward in transparent, explainable financial intelligence.

Built With

Share this project:

Updates

Submission history