KestrelCV
Flight-demonstrated TRL 5→6 Dual-use EU + NATO
The airframe is replaceable.
The perception layer is the product.

A portable onboard mission payload that adds real-time perception to standard UAV carriers — detecting, tracking and visualizing mission-relevant objects directly onboard, over the existing video link. No cloud. No platform lock-in.

scroll ↓
01 · Problem

A UAV can see the scene.
Holding what matters is still a human job.

The perception gap sits inside the loop — between a live feed and a confident, held target.

Operator load

Eyes on a laggy feed

The operator hunts, identifies and holds the object by hand over the video link — under stress, wind and a moving airframe. Miss a frame, lose the track.

Contested links

No cloud at the edge

Server-side AI is unusable where it matters: jamming, latency and lost link. The perception has to live on the aircraft, not in a datacenter.

Wrong AI

Off-the-shelf can't see from above

Generic detectors are ground-trained. Small, oblique, top-down objects slip through — and they false-fire on rooftops and field texture.

02 · Solution

What KestrelCV does

A portable onboard perception layer that rides the aircraft — detect many, track one, assist the operator, all onboard.

Detect many

A live sensor plot

A whole-frame detector marks every vehicle, person and mission-relevant object in view with a clean reticle — a live plot, not screen clutter.

Track one

Lock & hold

The operator — or the system — selects a single object; KestrelCV holds it through motion, vibration and scene change, with range and closing rate on the HUD.

Assist

Through the link they already have

The full perception overlay travels down the existing video link into goggles or ground station — the operator sees what the aircraft sees, decides and acts.

Onboard

No cloud, no lock-in

Runs headless on the airframe. Zero external services at inference — no continuous external-compute dependency.

Design doctrine

The selected target is priority #1. Occlusion → resume the same track, not a new detection. The reticle stays glued; the picture reads expensive — because perception is trust.

03 · Technology

Software, not a box

A portable perception engine. Low latency. Any edge.

~30ms
internal processing latency
2
compute backends validated
0
cloud calls at inference
1
video path · camera → operator
Portable

Runs anywhere

A hardware-agnostic software stack, validated on two very different compute backends. The same core scales onto more powerful or leaner silicon — the code doesn't change with the hardware.

Low latency

~30 ms onboard

Internal perception latency low enough for a human to fly on — and for the loop to close onboard. Tuned per compute load.

Through-channel

One video path

The perception overlay rides the existing video link — no second datalink, no cloud. The operator sees exactly what the aircraft perceives.

04 · Deployment

Three ways to fly

One fast video channel — the operator chooses how much of the loop to hand over.

Autonomous

Closed loop

Detect → lock → track with no human in the inner loop. Latency low enough that the aircraft holds the object through its own and the target's maneuver.

Semi-autonomous

Operator's co-pilot

The system finds and holds candidates; the operator designates and commits. Machine speed for perception, human for the decision.

Assisted

Augmented reality

A pure perception overlay painted into the goggles — the operator flies, the HUD calls out what matters, in real time, over the link they already have.

05 · Missions

One core, defence-first & dual-use

One perception core keeps its picture onboard — degrading gracefully on a contested link rather than going blind.

Scales across a family — one perception core; the platform picks its class. The software doesn't change, the mount does.

Light · FPV
Fixed camera

Fixed camera core + gyro-EIS.

Recon
Stabilized

1-axis stabilized + software EIS.

ISR
Gimballed

3-axis gimbal + geolocation.

06 · Traction

It already flies

Not slideware — a demonstrator in the air. It flew on a real airframe, locked a moving vehicle and held it — with the perception HUD carried live into the operator's goggles over the video link. Built by two founders, concept to flying portable demonstrator, without outside funding.

Whole-frame vehicle detection Live
Single-target lock & hold Live
HUD over the video link Live
Software-only image stabilization Live
Two compute builds (portable) Live
Repeatable field validation In progress
07 · Programme

From flight-demonstrated payload
to fielded capability

A structured 12-month engineering programme — turning a proven payload into an industrial-grade mission system.

WP0Done

Flight-demonstrated payload

Detect + lock + HUD, in the air, on two compute builds.

WP1In flight

Autonomy & co-pilot modes

Autonomous lock, co-pilot assist & AR overlay on one core.

WP2

Geolocation / navigation fusion

GPS + AHRS + camera angle + range → object lat/lon.

WP3

Maneuver-grade tracking

Hold through aggressive platform & target maneuver.

WP4

Full-motion stabilization

Gyro + optical-flow EIS across the platform family.

WP5

Advanced mission carrier

Video links, gimbal, compute carriers, airframe fit.

WP6

Hardening & endurance

Thermals, 24 h stress, anti-tamper, secure boot.

WP7

Repeatable field validation

Structured flight campaigns; labeled evaluation set.

08 · Team

Two senior founders — hardware & software

Decades of hands-on software, embedded, hardware and systems engineering — from the bench to the airframe. Build-first: the demonstrator flew before the pitch was written.

Co-founder · Software

Flight & perception

Flies the airframe and owns the perception core — detection, target tracking, video systems and the onboard HUD. Took the system from zero to a held target in the air.

Co-founder · Hardware

Electronics, carriers & integration

Owns the physical stack — electronics, video links, stabilization, mission-carrier design and compact payload integration onto the airframe.