Meridian
Real-Time Sub-500ms Geospatial Operations Engine
Role
Full-stack systems engineer: Rust WebSocket streaming, PostGIS spatial queries, ML anomaly pipeline
Primary Stack
Rust · PostGIS · Redis Pub/Sub · LangChain · WebSocket · React
“Track 200+ moving entities across a geographic territory in real-time, detect spatial anomalies automatically, and let analysts query the fleet in plain English instead of writing raw SQL.”
Sub-500ms latency end-to-end from entity coordinate update to dashboard render
ML anomaly model had to run CPU-only in constrained edge environments
Natural language interface had to be guarded against SQL injection
Why PostGIS over raw lat/lng math in SQL?▼
Geospatial queries on raw coordinates require expensive spherical trigonometry in SQL. PostGIS turns 'find all entities within 500m of perimeter X' into a single ST_DWithin() call backed by a spatial GIST index, yielding 10x faster lookups under high concurrency.
Why Rust Tokio for the streaming service?▼
Handling 200+ concurrent WebSocket streams with a sub-500ms latency SLA requires predictable execution. Rust gives zero garbage collection pauses and Tokio handles massive async fan-out with minimal memory footprint.
Distributed System Architecture
Explore the multi-tier topology below. Switch between the interactive blueprint canvas, standard Mermaid.js flowcharts, and the step-by-step request simulator.
High-frequency coordinate streaming engine in Rust with PostGIS spatial indexing, Redis Pub/Sub broadcast, and natural-language-to-SQL AST compiler.
Engineering Post-Mortem & Next Iteration
I would add a 5-second TTL cache for repeated geofence checks to cut PostGIS load by ~70% during peak operations.