▸ CASE STUDY 001

CLIMATE BOND RISK

Solving the Climate-Pricing Gap in Municipal Finance.

GEOSPATIAL AI PYTORCH FASTAPI SENTINEL-2

Most bond traders look at spreadsheets. I look at satellites.

Municipal bonds funding your local infrastructure are traditionally priced on tax bases and ledger sheets. But in California, a wildfire doesn't care about your tax base—it deletes it.

[ VIZ: SATELLITE OVERLAY HUD ]

The "Problem"

Standard credit models lag significantly in pricing physical climate risk. This creates a "Climate Spread"—unpriced risk that could lead to $100M+ portfolio impairments if not identified early.

The Engineering

I built a hybrid ML architecture to ingest and process 524k+ NASA FIRMS records. We didn't just guess; we used real physics. The system calculates NDVI (Vegetation Health) from Sentinel-2 multispectral imagery to measure how "ignitable" the landscape is around each bond asset.

Then, the fun part: A temporal CNN + LSTM model that looks at 7-day weather sequences to predict if a fire path is heading towards an asset. It's essentially a "radar" for financial risk.

[ VIZ: FIRE PATH TRAJECTORY ]

Why it Matters

We quantified an average 78 bps yield spread across a $1.2B portfolio. In institutional finance, 78 basis points is the difference between a safe bet and a systemic failure.

"Finally, a model that sees the fire before it hits the portfolio."