Skip to content
AI Games — software & AI studio, Sofia

PRODUCT · PRE-LAUNCH

Praemonitus.io

One coordination layer for a fleet of mixed drones, robots and unmanned vehicles. Software only, vendor-agnostic, built for civilian work — search and rescue, agriculture, environmental monitoring, logistics.

Type
Autonomous systems · platform
Status
Pre-launch — early-access list open
Hardware
ROS 2 + ArduPilot — no proprietary radios
Decisions
Bayesian probabilistic outcome simulation

What it is

Praemonitus coordinates civilian fleets — search-and-rescue teams, agricultural robots, environmental sensors, logistics rovers — through one map, one timeline and one decision surface. Underneath, it is a modular software layer over ROS 2 and ArduPilot with a Bayesian decision engine that scores risk and surfaces abort thresholds before the fleet commits.

Why we built it

There is no unified, vendor-agnostic coordination layer for civilian mixed-fleet operations today. Praemonitus is the studio's answer — fleet-level orchestration above, open VLA manipulation policies (SPEAR-1 by INSAIT on compatible embodiments) below, and a calibrated decision wrapper deciding when an autonomous step proceeds and when an operator is asked.

Stack

  • ROS 2 robotics middleware · ArduPilot ecosystem
  • Bayesian probabilistic outcome simulation engine for fleet-level decisions
  • On-robot manipulation policy: SPEAR-1 by INSAIT (Gemma license) on Franka Research 3 and WidowX
  • Custom fine-tuned LLMs for risk scoring and mission planning
  • Claude Sonnet 4.6, GPT-4.1, Gemini 2.5 Pro via OpenRouter / AI Gateway for fleet-level coordination
  • Hybrid edge / cloud inference — local GPU or cloud APIs

Building autonomous operations?