rotastellar
The operating system for orbital compute.
The operating system for orbital compute.
Satellites lose solar power every 90 minutes. Compute must be scheduled around eclipse cycles, battery reserves, and thermal limits.
Ground station passes last minutes and happen a few times per day. Data must queue, compress, and transfer across windows.
Link quality changes with elevation angle. Bit error rates fluctuate. Forward error correction must adapt dynamically per pass.
10–100 bit flips/device/day from radiation. Zero convection for cooling. Software must be autonomous and fault-tolerant.
Billions are going into hardware. Nobody is building the software layer. These satellites need an operating system that understands orbital mechanics.
Every orbital compute company retrofits cloud models to space.
We designed compute primitives that are native to orbital physics.
Eclipse periods aren't downtime — they're scheduled compute windows. Communication passes aren't constraints — they're deadline-driven execution units. Orbital mechanics isn't a limitation — it's the programming model.
Real-time tracking of 14,610 satellites. Orbital propagation, conjunction detection, feasibility analysis.
Constraint-aware scheduler. Decides what runs on-board vs. ground based on power, thermal, and data reduction.
Stateless SGP4/SDP4 propagation. Position, velocity, eclipse state, ground passes — digital twins on demand.
Pull-based protocol for satellites with intermittent contact. Autonomous execution with checkpoint/resume.
Plan, deploy, and monitor orbital compute. Constellation health, alerts, team collaboration.
pip install rotastellar to mission control dashboard.
Handles the physics so developers can focus on their workloads.
4-phase constraint-aware planner. Automatic on-board vs. ground placement with adaptive FEC selection. Published research paper.
Full SaaS product. Missions, plan comparison, real-time globe with 14.6k satellites, constellation health monitoring, team workspaces.
Users define custom satellites with orbital elements. Full monitoring: passes, visibility, latency analysis. Private by default, org-scoped.
Rust agent for on-satellite execution. Python SDK for developers. Pull-based protocol designed for intermittent ground contact.
10-factor satellite feasibility scoring: altitude, eclipse, thermal cycling, communication windows, power budget. Letter grade + recommendations.
Batch planning across entire constellations. Health aggregation: conjunction risk, feasibility scores, fleet-wide monitoring with alerts.
Stateless SGP4/SDP4 propagation at the edge. 8 endpoints for position, velocity, eclipse state, ground passes, trajectory, and constellation templates.
Pick a preset (ML inference, earth obs, federated learning) or build a custom step DAG with dependencies.
Choose from 14.6k tracked satellites, import from the rotastellar catalog, or define custom orbital elements.
CAE engine builds physics-grounded plan: placement decisions, transfer scheduling, window allocation.
Agent executes on-orbit. Real-time telemetry, event streaming, checkpoint/resume across contact windows.
SaaS subscriptions are the entry point. As customers move to production deployments, revenue expands through per-execution compute pricing (metered API usage), enterprise contracts for dedicated ground station time, and managed agent deployments for constellation operators.
First execution model designed for orbit, not adapted from cloud. Eclipse-steps run during darkness, window-steps adapt to deadlines, pass-steps fit single contacts. No prior art exists.
Every orbital scheduler returns one plan or fails. Ours returns a Pareto frontier of trade-offs: fastest vs. highest quality vs. lowest energy. Operators choose, not the system.
No system executes multi-step workflows across multiple satellites. We distribute DAG steps across constellation members with ISL routing and fault-tolerant partial results.
IoT checkpoints reactively on power loss. Orbital eclipses are predictable to the second. We checkpoint proactively before predicted hazards — zero wasted computation.
Cloud has vCPUs. Space has nothing. We define the first formal resource abstraction for heterogeneous satellite fleets — from 3U CubeSats to large buses.
The only platform from catalog to execution: tracking, planning, on-orbit agent, mission control, SDKs. Hardware companies build satellites — we build the software layer.
Rota, Inc. is a family of companies building infrastructure for AI and compute. rotastellar is the space computing division. Our team brings experience from Google, Microsoft, and leading aerospace companies, with backgrounds in data engineering, AI research, cloud infrastructure, and aerospace engineering.
Parent company based in San Francisco. Also operates rotalabs.ai (AI safety research) and rotascale.com (enterprise AI infrastructure).
Lean deployment on Cloudflare Workers (global edge, auto-scaling) + OVH dedicated server. Efficient infrastructure — entire platform runs on minimal compute.
Hire 2 engineers (orbital systems + distributed systems), secure enterprise design partners, and ship constellation DAG orchestration — the first multi-satellite workflow execution engine.
Try it now