Krishi-Kavach

Inspiration

In India, millions of farmers face devastating crop failures, but insurance payouts (PMFBY) are often delayed by months of manual paperwork. We built Krishi-Kavach ("Crop Shield") to replace bureaucratic delays with a "Parametric Shield"—triggering instant payouts by correlating a farmer's immediate distress reports with official satellite climate data.

What it does

It’s an automated parametric insurance pipeline. It validates farmer distress calls (IVR Voice) against official IMD satellite rainfall grids using a ±2-hour Temporal Join. If our confidence model (60% satellite, 40% voice) hits a 70% threshold, a payout is automatically simulated and visualized on a dashboard, bypassing all manual verification.

How we built it

We used the Databricks Data Intelligence Platform with a Medallion Architecture:

Bronze: Clean ingestion of raw IoT and IVR streams. Silver: High-performance PySpark engine for fuzzy temporal joins and confidence scoring. Gold: Aggregated payout simulations for district-level risk assessment. Frontend: A Streamlit Dashboard for real-time visualization of triggers and payouts.

Challenges we ran into

Synchronizing "subjective" farmer calls with "objective" satellite grids was the biggest hurdle. Farmers often report issues hours after an event. Implementing a Scalable Temporal Window Join in PySpark that could handle these "fuzzy" overlaps while ensuring data integrity was a significant technical challenge.

Accomplishments that we're proud of

End-to-End Automation: A zero-human-intervention insurance trigger system. Reliable Logic: A confidence model that fairly balances human reporting with hard scientific data. Visual Transparency: A dashboard that makes complex insurance logic intuitive for stakeholders.

What we learned

We learned the power of Delta Lake for handling real-time IoT updates with ACID guarantees. Most importantly, we realized that in agriculture, "ground truth" (satellite) is most effective when combined with "human truth" (farmer reporting).

What's next for Krishi-Kavach

Real-time Payments: Integrating directly with banking APIs for instant UPI payouts. Predictive Shield: Using ML to predict triggers before they happen. Mobile App: A one-click reporting app with automated geolocation for farmers.

Built With

• databricks
• delta-lake
• matplotlib
• pyspark
• python
• streamlit
• unity-catalog