Post

Swarm House — On-Prem Data Platform for Autonomous Drone Swarms

Design and runnable PoC for an air-gapped drone-swarm data platform: Parquet/DuckDB, peer sync, ground warehouse, simulation.

Swarm House — On-Prem Data Platform for Autonomous Drone Swarms

Engagement · 2025–2026 · role: platform / DevOps design (client undisclosed)
Repo: swarm-house on git.produktor.io

An on-prem, air-gapped data platform sketch for autonomous drone swarms: high-rate telemetry and on-board detections stay local; only compact derived state crosses the mesh; full flights offload to a ground warehouse after landing.

What was delivered

AreaIn-repo
Design docs + ADRsProblem → architecture → sync → security → CI/CD (docs/)
Executive + journey00 — Executive · 12 — Design journey
SimulatorVirtual fleet writing Hive-partitioned Parquet / DuckDB
Prototypes2D (and 3D WIP) swarm visualization — React / TypeScript / Vite
Infra PoCAnsible host prep · Terraform sim/ground · Flux overlays · k3d warehouse with MinIO

Design principles (short)

  1. Every drone is autonomous — no swarm-wide orchestrator over intermittent mesh.
  2. One storage layout on drone, warehouse, and bench (Parquet + DuckDB).
  3. Sync results, not raw high-rate telemetry, while airborne.
  4. Zero-trust keys provisioned ahead of deployment.
  5. SQL as the peer query contract (read-only gate / forced-command path).

Client identity is omitted on purpose; this page describes the technical work only.

This post is licensed under CC BY 4.0 by the author.