Counter-swarm intelligence · SDK · pre-launch

Track the swarm.
Find the invisible.

FalconEye is the AI brain for counter-swarm defense — tracking whole drone swarms in real time and detecting the RF-silent autonomous drones that every existing CUAS system misses. The SDK, not the system.

RF-silent detection Sensor-agnostic Swarm-scale tracking
RADAR · LIVE
THREATS 0 RF-SILENT 0 BIRDS REJ. 0

The brain, not the box.

We don't build the system. We build the one software layer that's underbuilt and rising in value — the intelligence that sits above any sensor and below any kill chain — and license it to the people who own the rest.

What it does

Three hard problems. One SDK.

FalconEye solves the three layers of the counter-swarm intelligence problem that sensor-level CUAS leaves unaddressed.

GM-PHD Swarm Tracker

Gaussian Mixture Probability Hypothesis Density filter tracks whole swarms without hard assignment — no ID thrash as drones close in on each other. Birth suppression, Euclidean merge, and persistent Hungarian labelling prevent the ghost tracks that trigger false alerts.

Benchmark: −14% GOSPA vs GNN baseline · 20-drone sim · 1 km radar

RF-Silent Kinematic Classifier

GPS-denied autonomous drones emit no RF and ignore GPS jamming — they navigate by vision. FalconEye detects them by their kinematic fingerprint: path straightness, heading consistency, and velocity smoothness read straight off the track — no RF, no GPS correlation.

Accuracy: 100% (24/24) · VIO drone vs GPS drone vs bird · 8-seed Monte Carlo

Distributed T2TF Fusion

Two geographically separated sensor nodes see the same swarm from different angles. Their error ellipses are complementary. Covariance Intersection fusion — the only mathematically consistent method for correlated estimates — merges them into a single site picture without double-counting shared history.

Position variance: −28% vs single node · 4,518 matched track pairs
3,893 lines of Python 59/59 tests passing Radar · EO/IR · Acoustic · any combination CoT/ATAK output to any C2 No new hardware required
Where it sits

Sensor-agnostic above.
C2-agnostic below.

In a modern counter-drone stack, sensor hardware and command systems are mature. The middle layer — the intelligence that fuses, tracks, and classifies — is still bolt-on afterthought.

FalconEye fills that gap. A software SDK that adapts to any radar, EO, or acoustic sensor above and emits standard Cursor-on-Target tracks below. Swap a sensor; nothing above the adapter changes. Switch to a different C2; the output format doesn't move.

This is the ARM model: license the brain, own the rest.

Any C2 / ATAK / command picture CoT · Remote ID
Threat prioritisation ranked watch-list
FALCONEYE SDK · detect · track · fuse · classify
Sensor ingest adapters radar · EO · acoustic
Any sensor hardware customer's existing kit
The insight
We detect the drones nobody else can — by how they move, not what they emit.

FALCONEYE · FOUNDING THESIS

Current CUAS is built around RF: find the command link, jam the GPS, intercept the pilot. None of that works against a GPS-denied autonomous drone. It has no command link to find. It uses no GPS to jam. There is no pilot signal to cut.

The only handle left is what the drone does — how it moves. Vision-guided autonomous flight leaves a distinctive kinematic signature, and that signature is exactly what FalconEye is built to read.

FalconEye · Result FalconEye tracking and classifying drones in a simulated scene
FalconEye SDK · Counter-UAS · Delaware C-Corp

The kinematic fingerprint detector

GPS-denied drones have a characteristic flight pattern: smooth velocity with periodic correction events, high path straightness, consistent heading. FalconEye reads that fingerprint from any radar or EO track. No RF needed. No GPS correlation needed.

Working SDK · pre-launch

Real code. Real tests. No vaporware.

FalconEye is a fully shipped MVP — packaged Python SDK with a complete test suite and four live demos. All metrics are simulation-validated; real-sensor field validation is the next step.

3,893
Lines of Python
Production-structured, typed, documented
59/59
Tests passing
Full pytest suite, zero failures
4
Core modules
Tracker · Classifier · Ingest · Fusion
4
Live demos
Single-site, RF-silent, T2TF, ingest replay
GM-PHD · −14% GOSPA vs GNN baseline RF-silent classifier · 100% accuracy · 8-seed MC CI fusion · −28% position variance · 4,518 pairs CoT/ATAK output · plugs into any C2 today

All benchmarks simulated on synthetic data. Real-sensor field validation is the primary use of pre-seed funds.

See it run

Watch the SDK in action.

A walkthrough of FalconEye tracking a swarm, rejecting birds, and flagging the RF-silent drone by its kinematic fingerprint — running live against simulated sensor data.

Who it's for

Built to be licensed, not sold as a system.

FalconEye is an SDK component. You own the kill chain, the certification, and the customer relationship — we provide the swarm intelligence you'd rather not build.

CUAS OEM vendors

Add the brain you don't have

Drop a swarm tracker and RF-silent classifier into your counter-drone platform without standing up a computer-vision or tracking team. Differentiates your hardware on the capability that's rising in demand.

Defence primes & integrators

Multi-sensor fusion, solved

Building a fixed-site CUAS with radar, EO, and acoustic? FalconEye's distributed T2TF layer fuses them correctly — Covariance Intersection, not naive Kalman — and emits a single coherent site picture.

Sovereign programs

A licensable intelligence core

Indigenous counter-drone capability that doesn't depend on a foreign prime owning your detection stack. Source-available licensing available for qualified sovereign partners.

Request access

Request the technical briefing.

We're in pre-launch and working with a small number of design partners. Tell us who you are and we'll share performance data, the integration spec, and a path to a pilot — under NDA, with qualified partners.

We reply to qualified partners within two business days.

Request received. We'll review and follow up at the email you provided. Nothing further is needed from you right now.