PATYX — Autonomous Rocket Systems

01 / Problem

The sky is full of
$50K drones.

Militaries are trapped between two bad options. FPV interceptors can't reach altitudes over 2 km or catch Shaheds at 500 km/h. Patriot-class missiles can't economically justify a $1M shot on a $50K target.

PATYX closes the gap. Faster than an FPV. Cheaper than a missile by 200×.

200×

Cheaper than a Patriot missile

2.4km

Altitude FPVs can't reach

500km/h

Target speed FPVs can't catch

02 / Product

Model S.
Anti-Shahed
interceptor.

High-speed smart rocket. AI terminal guidance. Launched from a portable station, vectored at subsonic airborne targets. Built to be produced in volume, deployed alongside existing mobile anti-drone groups.

IN DEVELOPMENT MVP Q2 / 26 BASE UNIT $5,000

FIG 01 · MODEL S / SCHEMATIC · REV 04.26 SCALE 1:8

Length

1000mm

Diameter

60mm

Launch mass

3kg

Payload

300g

Top speed

M 0.7850 km/h

Apogee

2.4km

Range

1–8km

Engagement

<10sec

03 / AI Pipeline

Design, test,
ship.
In days.

PATYX's moat isn't hardware. It's the AI brain that designs, simulates, trains, and iterates the rockets itself. Parametric CAD → physics sim → RL flight training → CAM handoff. Eight stages, Go/No-Go at each gate, zero manual handoffs.

Every rocket deployed is a telemetry node. Every flight trains the next fleet. OTA updates roll improvements to deployed systems in hours, not months.

[GO]

STAGE 01

Specification

Target class, envelope, thermal + RF constraints declared as parameters.

[GO]

STAGE 02

Parametric CAD

CadQuery generates geometry. Versioned. Diff-able. Reproducible from spec.

[GO]

STAGE 03

Physics sim

CFD, thermal, structural. Failure modes surface before hardware exists.

[GO]

STAGE 04

RL flight training

PPO agents train against synthetic Shahed traces. ONNX-exported policy.

[GO]

STAGE 05

Flight control

ArduPilot + MAVLink. Policy runs on-board. No dependency on operator.

[GO]

STAGE 06

Manufacturing

CAM post-processors from the same CAD. Part IDs back-propagate to design.

[GO]

STAGE 07

Field telemetry

Every flight logs to the fleet. Engagement outcomes feed the next training run.

[GO]

STAGE 08

OTA update

New policies deployed to fielded units. Hours, not months.

patyx.brain — build 04.26 SSH / 127.0.0.1

$ patyx compile --target model-s --optimize thrust-to-weight --autonomous

[GO] Stage 01/08 · Specification parsed · 34 parameters locked

[GO] Stage 02/08 · CAD generated · 1000mm × 60mm / 16kg

[GO] Stage 03/08 · CFD converged · Cd=0.41 @ M0.7 · thermal nom

[GO] Stage 04/08 · PPO converged · 2.4M timesteps · 94.7% hit rate

[GO] Stage 05/08 · Flight controller staged · ArduPilot v4.5

[GO] Stage 06/08 · CAM post complete · manifest 847 parts

[REVIEW] Stage 07/08 · Awaiting field telemetry · MVP deploy pending

> 8 stages. 8 Go/No-Go gates. 0 manual handoffs.

> From spec to flight-ready hardware in days._

04 / Market

The gap is
the market.

Every country defending against cheap attack drones needs a class of interceptor that doesn't exist yet. PATYX fits between FPVs and missiles on every axis that matters — altitude, speed, unit cost, production volume.

System

Unit cost

Speed

Altitude

Autonomous

FPV interceptor drone

~$2K

200 km/h

<1.5 km

Operator-piloted

Patriot (PAC-3)

$4M+

Mach 4+

24 km

Semi

IRIS-T SLM

$1.4M

Mach 3

20 km

Semi

Stinger MANPADS

$400K

Mach 2.2

3.8 km

Operator

PATYX Model S

$5,000

Mach 0.7

2.4 km

Full autonomy

Go-To-Market

Embed PATYX into existing mobile anti-drone groups. They already have trucks, radar, crews. We add interceptor reach + speed they lack today.

Business Model

$5K base rocket + control modules + launch stations. Recurring revenue from OTA AI updates as threats evolve on the battlefield.

05 / Traction

Past idea.
Into hardware.

Development status · Q2 / 26 5 / 6 COMPLETE

[GO] COMPLETE

Core modules

Guidance, propulsion, seeker, control assembled and integrated.

[GO] COMPLETE

Autonomous sim flights

Full engagement cycle validated. Physical reactions confirmed.

[GO] COMPLETE

Model S design

Geometry locked. 847 parts. CAM post complete.

[GO] COMPLETE

AI trained

PPO convergence at 94.7% simulated hit rate. ONNX-deployed.

[GO] COMPLETE

Ground testing

Propellant burns, seeker validation, thrust characterization complete.

[HOLD] NEXT

Battlefield MVP

Live engagement against real Shahed targets. Scheduled Q2/26.

07 / Team

Defense + AI
operators.

Business Development

Yura

Took the Plastun ground drone from concept to codified small-scale serial production.

CTO

Sergii

Built and deployed AI defense products. Author of a battlefield auto-targeting module.

COO

Roman

Ran 100+ FTE IT programs. Operator network across Ukrainian armed forces.

Sky
superiority
at $5K/shot.

The sky is full of
$50K drones.

Model S.
Anti-Shahed
interceptor.

Design, test,
ship.
In days.

The gap is
the market.

Past idea.
Into hardware.

Model S. Then M. Then L.

Defense + AI
operators.

Raising $250K
to battle-test
Model S.

Skysuperiorityat $5K/shot.

The sky is full of$50K drones.

Model S.Anti-Shahedinterceptor.

Design, test,ship.In days.

The gap isthe market.

Past idea.Into hardware.

Model S. Then M. Then L.

Defense + AIoperators.

Raising $250Kto battle-testModel S.

Sky
superiority
at $5K/shot.

The sky is full of
$50K drones.

Model S.
Anti-Shahed
interceptor.

Design, test,
ship.
In days.

The gap is
the market.

Past idea.
Into hardware.

Defense + AI
operators.

Raising $250K
to battle-test
Model S.