Rocket Platform
Integrated Environment for building AI-powered welding automation and inspection.
Rocket Platform combines industrial cameras, edge compute units, and embedded software to enable AI-powered welding automation directly at the source.
Rocket Neuron - Industrial processing unit
-
welding speed up to 1200 mm/min with 0.5 mm prec.
-
latency: 100–150 ms input video | 50–400 ms output signal
-
Industrial enclosures, DIN/19" rack mountable
-
protocols: OPC UA, MQTT, Modbus, ADS
-
HDMI and USB for industrial monitor (that can be also provided 21'5, hdr)
-
Browser-based video feed with AI overlays for additional inspection
Compute units are in variants to handle single camera input, 2 and also for 8 cameras.
Included Software:
-
provides pipeline designer, output signal designer
-
manages local recording and events
-
provides configurable preview with AI overlay
Rocket Trainer
RocketTrainer is the core of your automation workflow—train, simulate, and deploy AI models that talk directly to your machines.
Train and update AI models—quickly, no data science required.
RocketTrainer is a platform for building vision-based AI models that guide robots or check weld quality. It’s built for production engineers—no AI background needed. 😉
With RocketTrainer, you go from process video to a working model in hours—while staying fully in control of what the AI learns and how it behaves.
Why It’s Different?
Smart Dataset Management
Track exactly what data was used for each model version,
and ensure your dataset stays diverse and complete.
AI-Assisted Labeling
Use previous models to auto-label new footage—just review and correct.
No more tagging everything by hand.
Simulation Before Deployment
Test how your model performs on historical recordings before going live. Think of it like regression testing for AI.
One-Click Deployment to RocketWelder
Deploy models directly to your camera + edge compute unit. No extra tools. No manual steps.


Solutions:
Orbital Welding with a Positioner
Automate tank welding—even when the path drifts
Rotating a tank on a positioner gives you precision motion—but not a perfect seam. In real-world production, the weld joint often rises and falls slightly along the cylinder’s axis as it rotates.
Even 1–3 mm of axial drift can throw the torch off target—especially in TIG. The result? Missed joints, weak fusion, or inconsistent beads. Traditional robots can’t adapt, so skilled welders are often required just to follow the seam.

Why Fixed Robots Struggle
-
Seam doesn’t stay in a flat plane during rotation
-
Preprogrammed torch paths miss the actual joint
-
Minor variations create major defects
-
Laser profilometers often fail on reflective surfaces like aluminum
How Rocket Welder Solves It
RocketWelder tracks the weld seam in real time using a high-dynamic-range welding camera (Retina) and edge AI compute (Neuron). As the tank rotates, the system adjusts the torch vertically (Z-axis) to follow seam elevation—keeping the arc centered on the joint.
🎯 Live Seam Tracking
Robust to arc light, spatter, fume, and aluminum glare
📐 Z-Axis Torch Correction
Compensates for up to ±30 mm seam drift without re-teaching
🔁 One program – multiple tanks
The system guides the torch along the tank edge – changes in tank height or diameter do not require any adjustments to the system.
Clear overlays and no-code control—no welding certification required
Works With
-
Example of tank diameters: 20–60 cm / 8–24″
-
Welding types: TIG and MIG/MAG
-
Typical precision: ±0.5 mm at 1000 mm/min travel speed
Key Benefits
-
Enables automation where fixed-path robots fail
-
Empowers operators to run production without being certified welders
-
Handles aluminum and reflective materials better than laser-based tracking
-
Speeds up setup, tolerates real-world part variability
-
Keeps all data and AI models fully on-site



Full Automation for Long Structural Welds
Fully automated welding—no manual correction needed.
Why Traditional Systems Fail
Welding long components—like crane columns, beams, or frames—often involves:
-
Carriages that follow rails or machine references
-
Gantry arms manually adjusted by operators
The problem? These systems don’t track the actual weld seam—they follow external geometry or require visual corrections.
As the weld progresses over several meters, even small deviations in joint alignment or part straightness cause the torch to drift off the seam.
That’s why:
-
Carriages produce off-center welds unless the part is perfectly fixtured
-
Gantry arms depend on operators constantly correcting the position by hand
-
Tolerance stack-up over long distances leads to inconsistency and rework
How RocketWelder Solves It
RocketWelder makes long welds fully autonomous. It uses real-time vision and AI to detect the actual seam throughout the entire run—adjusting the torch to stay on track without human intervention.
🧠 Live seam detection using Retina welding camera
🔄 Automatic left-right correction during motion—no more manual nudging
🎮 Operator just loads the part and presses start
Whether mounted on a carriage or gantry, the system ensures consistent alignment from start to finish.

Works With
-
Example of part lenghts: 2–11 m / 6.5–36′
-
Welding types: MIG/MAG
-
Typical precision: ±0.5 mm at 1000 mm/min travel speed
Key Benefits
-
Enables full automation without operator correction
-
Handles real-world part variability and fit-up tolerances
-
Works on aluminum and reflective surfaces—no laser failure
-
Reduces rework and improves weld consistency
-
Cuts labor cost and boosts throughput
-
Keeps all model logic and video data on-premises
Welding Training & Certification Station
Soon on Kickstarter: AI-assisted feedback for the next generation of welders
Built for welding instructors, vocational schools, and certifiers.
Why Traditional Training Falls Short
Instructors can’t follow every weld in real time—and post-analysis lacks detail.
Mistakes are hard to catch and explain. There’s no objective record. And scaling requires more instructors, not just more booths.
How RocketWelder Solves It
RocketWelder turns a training booth into an AI-enhanced station.
It records each weld with high-fidelity video and overlays, tracking the torch, seam, and arc behavior live. Students get visual feedback that accelerates learning. Instructors can review welds later and grade with evidence.
💡 Launching now on Kickstarter
This is your chance to bring AI-powered weld tracking to your program—whether you're teaching beginners or certifying professionals.
Works With
-
Standard booths (100×100 cm / 3×3′ and larger)
-
TIG and MIG/MAG setups
-
Manual training or certification tests
-
On-premise or mobile setups
Key Benefits
-
Provides real-time visual feedback for faster skill development
-
Makes training scalable—fewer instructors, more results
-
Records every weld with objective, repeatable metrics
-
Students can share branded recordings to promote their skills and instructors
-
Keeps data private—stored securely on-site