Liquid Edge
Team led by founder/CEO Vikash Kodati (IIT Guwahati, Int'l Space University), 20+ years at T‑Mobile/Capital One, building AI edge geospatial and cloud-native satellite-imagery systems.
Project Description
Edge devices, sensors, or lightweight models often emit streams of detections without an easy way to turn those signals into real-time situational insight.
What It Does:
Edge Event → Hotspot Relay is an event-driven backend built entirely inside the existing LiquidMetal MCP environment. It ingests geospatial detection events, aggregates them into location-based “hotspots,” and automatically issues alerts when activity crosses defined thresholds.
Core Features:
/api/ingest — HTTP endpoint for incoming detection events (lat, lon, object_type, confidence, timestamp).
SmartBucket edge_events stores raw inputs.
Trigger + SmartSQL hotspot_counts aggregates events by geohash and object type.
Secondary trigger creates SmartBucket alerts when count > 10 within 5 minutes.
/api/hotspots and /api/alerts expose live summaries as JSON.
Demo / Runtime Instructions:
Run inside existing LiquidMetal MCP runtime (no new server needed).
Post sample events:
curl -X POST /api/ingest -d ‘{“lat”:47.61,”lon”:-122.33,”object_type”:”ship”,”confidence”:0.9,”timestamp”:”2025-11-08T19:00Z”}’
View results:
curl /api/hotspots
curl /api/alerts
Judging Criteria:
Creativity — demonstrates event-driven geospatial analysis within a serverless MCP.
Reliability — uses LiquidMetal’s SmartBuckets/SmartSQL for atomic storage + triggers.
Real-World Impact — pattern applies to monitoring, logistics, and rapid-response systems.
Technologies Used:
LiquidMetal MCP server (primitives: sources, SmartBuckets, SmartSQL, triggers, tools); Python for trigger logic and event generator; curl for testing.
Scalable and fault-tolerant by design through MCP’s stateful runtime and trigger automation.
Prior Work
No prior code base; entire application built during the hackathon using LiquidMetal MCP runtime.