TERVOX GRID

The Threat

Drones are a growing security threat across NATO's eastern flank

Commercial and modified UAVs increasingly probe military and civilian infrastructure — and today's detection systems have structural blind spots.

The Solution

Detect, classify, and track — without cloud or cellular

Tervox Grid is a distributed sensor mesh that detects, classifies, and tracks drones using multi-modal AI fusion — fully autonomous at the network edge.

👁 Visual

Edge AI object detection (YOLOv8) on dual cameras with Hailo-8L neural accelerator. Real-time classification at 30+ FPS.

🔊 Acoustic

Microphone array with ML-based drone audio signature detection. Works in zero-visibility conditions — fog, night, smoke.

📻 Radio Frequency

Passive RF monitoring: Wi-Fi Remote ID reception, ADS-B aircraft context filtering, spectrum baseline monitoring.

▼ MULTI-MODAL FUSION ENGINE ▼

Portable, weather-hardened sensor node — rapid perimeter deployment

How It Works

From perimeter sensor to command & control

Four layers, one resilient pipeline. Sensor nodes talk over an encrypted LoRa mesh — the perimeter needs no cellular coverage at all.

Sensor nodes communicate via encrypted LoRa mesh — no cellular dependency on the perimeter

Why Tervox Wins

Built for the environments where others fail

Cloud-dependent competitors assume a permissive RF environment. Tervox Grid assumes the opposite.

Progress

Actual development progress

This section tracks the real state of the system as it is built — validated milestones, work in integration, and what comes next.

SENSOR NODE SUBSYSTEM

IN INTEGRATION

Edge detection platform built on Raspberry Pi 5 with Hailo-8L neural accelerator (13 TOPS INT8). The trained YOLOv8n drone model runs on-device; the acoustic array and RF receivers extend detection beyond line of sight.

• Edge AI vision: YOLOv8n compiled to .hef, real-time inference on Hailo-8L at 30+ FPS.
• Multi-sensor framework: extensible architecture for video, audio, and RF detection channels.
• ADS-B context layer: RTL-SDR + dump1090 aircraft tracking for false-positive filtering.
• Reproducible provisioning: Pi Imager → Ansible → Docker Compose, three-stage node bring-up.
• Telemetry isolation: MQTT bridge keeps the internal mesh unexposed.

C2 GATEWAY & OPERATIONS CONSOLE

OPERATIONAL

Fully functional MVP. Ingests sensor telemetry over mutual-TLS MQTT, persists state to PostgreSQL, broadcasts Cursor-on-Target (CoT) multicast for ATAK integration, and streams live updates to an offline-capable web operations console.

• Secure ingestion: MQTT over mTLS (TLS 1.3) with embedded PKI for node commissioning.
• Tactical broadcasting: CoT XML over UDP multicast — plugs into ATAK and existing C2 systems.
• Live operations console: MapLibre GL with offline vector tiles, real-time node markers, directional threat vectors, geofenced perimeters.

• Authentication: OIDC via Keycloak (Authorization Code + PKCE), HTTPS end-to-end.
• State & audit: PostgreSQL 16 persistence; all node lifecycle transitions logged for incident reconstruction.
• Test coverage: Testcontainers integration suite validating the full ingestion → broadcast flow.

STACK: Java 21 · Quarkus · PostgreSQL 16 · Eclipse Mosquitto (MQTT mTLS) · Keycloak OIDC · MapLibre GL JS · Docker

Market Opportunity

A rapidly expanding counter-UAS market

Roadmap

From field PoC to series production

Founder

Deep-tech experience meets mission-critical engineering

Partnership Opportunities

How you can get involved