Raamenomics

Large hedge funds have tools to do risk assessment but the average consumer only has a stock chart. We provide mathematically backed models paired with LLM analysis to do multi-layer risk tests.

Comment

Inspiration

We were inspired by a core gap in retail and even many institutional portfolio tools: they show what moved, but not how risk propagates through a connected system. Markets behave like networks, not isolated tickers. A supply chain disruption, policy change, or geopolitical event can cascade across sectors and exposures in non-obvious ways. We wanted to make that systemic behavior understandable and interactive.

What We Built

We built Raamenomics, a full-stack portfolio risk intelligence platform that lets users upload a simple holdings CSV (company, shares) and immediately run multi-layer risk analysis across:

  • City-level structural risk
  • Graph-based dependency propagation
  • Macro scenario simulation
  • Third-order emergence analysis

The platform combines deterministic algorithms with real-world data context and AI-generated explanations.

How It Works

  1. User uploads a portfolio CSV.
  2. Backend normalizes holdings into weighted positions and computes value.
  3. Holdings are mapped into sector/infrastructure/meta-risk layers.
  4. Shocks propagate through a weighted dependency graph.
  5. Stability and tail-risk metrics are computed.
  6. Backboard generates portfolio-specific narrative explanations.
  7. Results are visualized across city, graph, simulation, and emergence tabs.

Tech Stack

  • Frontend: React, TypeScript, Vite, Tailwind, Framer Motion, React Three Fiber
  • Backend: FastAPI (Python)
  • State/Data Layer: Zustand + server-side analytics
  • AI Layer: Backboard API for scenario explanations and dynamic scenario generation
  • Data Sources: Live quote/news context from public APIs and RSS feeds
  • Deployment: Vercel-compatible frontend + API configuration

Key Capabilities

  • CSV-driven portfolio ingestion
  • Live portfolio valuation and weighted exposure
  • Sector/infrastructure/environment risk layering
  • Graph shock propagation with causal chains
  • Dynamic macro scenarios from real-world news flow
  • Third-order systemic emergence diagnostics
  • AI explanations tied to user portfolio and current context

Mathematical Model

For full mathematical details and implementation references, please refer to our GitHub repository.

Challenges We Faced

  • Keeping all tabs synchronized to the latest uploaded CSV without stale state
  • Balancing deterministic quant outputs with AI explanation quality
  • Ensuring graph layout and propagation remained readable at scale
  • Handling API reliability/fallbacks for real-time data and LLM responses
  • Preparing robust deployment routing for frontend + backend under one host

What We Learned

  • Deterministic models are essential for trust, but users also need interpretability.
  • AI is most useful as an explanation and scenario-generation layer, not as a replacement for core risk computation.
  • Systems thinking (dependency graphs, cascades, loops) gives a much stronger risk picture than isolated factor views.
  • Production-readiness requires as much work on state consistency and deployment configuration as on model design.

Why This Matters

Raamenomics reframes portfolio analysis from static performance tracking to systemic risk intelligence. Instead of only asking “what dropped?”, users can ask:

  • What is the shock?
  • Where does it propagate next?
  • Which holdings are structurally exposed?
  • How does emergence risk evolve as conditions change?

That is the shift we set out to build.

Built With

Share this project:

Updates