What Comes Back to Claude Stats Only, Never Raw Data

EMR & Spark isolated in Private VPC • Claude connects via SSH Tunnel • Stats returned through tunnel

Phase 1: Ingesting Raw Rows | 0.0s / 8.0s
1. Private VPC Enclosure EMR + Spark

EMR cluster, Spark executors, and raw Iceberg tables remain locked within an isolated VPC with no public IP and zero external ingress.

2. Secure SSH Tunnel AES-256

Claude Code connects strictly via an encrypted SSH tunnel. Job definitions dispatch inwards across the bastion boundary.

3. Returned Through Tunnel Stats Only

Spark aggregates data in VPC memory. Only 4 scalar statistical metrics are routed back through the SSH tunnel to Claude.