RiskRadar

Institutional OSINT surveillance for corporate lenders. Identify emerging legal, regulatory, and operational risks in real-time, months before traditional financial reporting lags.

Comment

Inspiration

Traditional credit monitoring is a "lagging" process; lenders often don't know a borrower is in trouble until they receive a quarterly financial statement that is already months old. I was inspired to build RiskRadar to solve this "information gap." By focusing on OSINT (Open Source Intelligence), the app identifies leading indicators of distress—such as legal disputes, management shifts, or operational ripples—that appear in the public domain long before they hit the balance sheet.

What it does

RiskRadar is a professional surveillance terminal for corporate lenders. It uses real-time search grounding to scan the web for non-financial risk signals. Key features include:

  • Covenant Mapping: Automatically analyzing news to see if it triggers "Technical Defaults" or "Material Adverse Change" (MAC) clauses.
  • Supply Chain Ripples: Visualizing how distress at one node (a supplier or customer) might impact the borrower.
  • Deterministic Proof: To ensure institutional trust, every signal is grounded with a "Source Proof" link, allowing officers to verify the original news or filing instantly.

How I built it

The application is built using React 19 and Tailwind CSS for a high-density, enterprise-grade UI. The core intelligence engine is powered by the Gemini API.

  • I utilized Search Grounding to ensure the model has access to current 2025/2026 data.
  • I implemented a "Bring Your Own Key" architecture to ensure institutional data privacy.
  • I built a custom JSON Extraction Layer to handle structured data output from the model when combined with real-time web tools.

Challenges I faced

A significant technical challenge was a 400 INVALID_ARGUMENT error encountered when attempting to use responseMimeType: "application/json" simultaneously with the googleSearch tool. Since the API does not currently support these together, I had to develop a robust manual parser to extract and validate JSON structures from the model's natural language stream, ensuring the app maintained its structured data integrity without losing real-time search capabilities.

What I learned

I learned that in Fintech, "Explainability" is more important than "Prediction." Lenders don't want a "black box" score; they want the actual source link. This project taught me how to shift from generative AI (creating content) to Surveillance AI (summarizing and grounding facts).

Built With

  • data
  • google-gemini-api-(@google/genai)
  • osint
  • proxies
  • react-19
  • search-grounding
  • tailwind-css
  • typescript
Share this project:

Updates

posted an update —

Project Overview: RiskRadar

Category: Keeping Loans on Track

Project Description

RiskRadar is a professional, institutional-grade surveillance terminal for corporate lenders. It bridges the “Information Gap” in credit monitoring—the dangerous window between quarterly financial reporting cycles.

While traditional credit monitoring relies on lagging indicators (financial statements that are often 3–4 months old), RiskRadar identifies emerging risks in real time using Open Source Intelligence (OSINT) and search-grounded AI. It surfaces material adverse events—such as litigation, regulatory delinquency, and operational disruptions—before they impact a company’s balance sheet.

Key Features

  • Real-Time Risk Surveillance Feed
    Continuously monitors the digital domain for non-financial risk signals. Signals are categorized into:

    • Legal
    • Regulatory
    • Management
    • Operational
      Each signal is severity-scored from Critical to Low.

  • Covenant Mapping
    Automatically maps public risk signals to common credit agreement clauses. Identifies whether events may trigger:

    • Material Adverse Change (MAC) clauses
    • Key Man departures
    • Technical Defaults

  • Supply Chain Ripple Analysis
    Visualizes second-order risks. If a borrower’s critical supplier faces distress, RiskRadar assesses the potential ripple effect and contagion risk to the borrower’s cash flow.

  • Internal Assessment Memo Generator
    Instantly drafts institutional-grade credit memos for Credit Committee review, synthesizing all observed signals into a clear, professional narrative.

  • Zero-Hallucination “Source Proof”
    Every risk signal is accompanied by a clickable URL and metadata sourced via Google Search Grounding, ensuring all claims are auditable and verifiable.

Target Users

  • Commercial Banks
    Automated Early Warning monitoring for mid-market loan portfolios.

  • Private Credit Funds
    High-touch surveillance of private investments where public data is often the only real-time signal.

  • Asset & Portfolio Managers
    360-degree monitoring of distressed debt and turnaround situations.

  • Credit Officers & Legal Teams
    Instant identification of technical default triggers and material adverse changes without manual research.

Tech Stack

Frontend

  • Built with React 19, TypeScript, and Tailwind CSS
  • High-density Terminal-style UI optimized for professional financial environments

Intelligence Engine

  • Powered by Google Gemini 3 Pro and Gemini 3 Flash

Search Grounding

  • Integrates the @google/genai Search Grounding tool
  • Provides access to real-time 2025/2026 data for reasoning about current corporate events

Deterministic Logic Layer

  • Custom extraction layer to address the current API limitation where
    responseMimeType: "application/json" cannot be used simultaneously with search tools
  • Preserves structured data integrity without sacrificing live web access

Deployment Architecture

  • ESM-based module structure
  • Designed for high performance and compatibility with modern edge-deployment platforms

Trust & Compliance

RiskRadar is built on a Privacy-First OSINT Protocol.

  • Does not require or analyze private borrower data
  • Fully compliant with strict banking privacy perimeters
  • Focuses exclusively on public signals, ensuring lenders remain the most informed party in any credit relationship

Log in or sign up for Devpost to join the conversation.