Ukraine
Decoy discrimination
- Problem
- Amplitude-only can’t tell real assets from decoys.
- Phase information
- Phase stability exposes the material: metal vs rubber.
- Operational output
- Fewer false positives, fewer wasted resources.
Our purpose
PerspeQtive builds AI for physical signals
We turn raw radar into mission-grade intelligence.
In radar, the richest information lives in the phase and wave structure, before the signal ever becomes an image. Standard computer vision throws it away. We read the full wave.
What we build
AI for physical signals
Where we start
Raw radar / SAR
What you get
Mission-grade intelligence
AI's current blind spot
The most valuable radar data is discarded by AI models
Radar data contains two signals: amplitude and phase. Current AI models only learn from amplitude, and throw phase away.
Sensor output: full radar data
Full signal: amplitude + phase captured
The AI bottleneck
Phase is discarded: 50% data loss
Blind decisions
Amplitude-only models fail: missed targets, false alarms
50%
Raw data thrown away
M$
Lost at the software layer
The lost information, visually
When radar becomes a black & white image
Most AI models treat phase as noise and discard it, becoming blind to valuable physical signals.
What gets destroyed
Phase texture · material signature · millimetric variations
Same scene, two readouts
Drag to see what the phase recovers
Left: what an amplitude-only model learns from. Right: the same return read with the full wave. Same sensor, same pixels — the difference is the phase.
Full wave · PerspeQtive Amplitude only The core
An engine that reads the full wave
Our engine reads amplitude and phase on standard GPUs: the part of the signal others discard. It's built and validated on real SAR today, and it's the core we build our radar models on. No quantum hardware.
Your raw SAR archive
Full-wave engine
Models that see more
Our edge is mathematical. We use quantum mathematics to represent wave structure that image-based AI can't, running at AI speed on standard GPUs.
Runs on standard NVIDIA GPUs No quantum hardware First patent being filed Pre-incubated · Quantum Launchpad
Why now · where it matters
Phase unlocks what amplitude-only AI can't see
Across today's defining theaters — Ukraine, the Middle East, Europe's contested seas — the easy signal (transponders, GPS, clean imagery) is failing or being spoofed. Here's what reading the full wave makes possible, and the missions where it matters.
The Middle East
Millimetric perturbations
- Problem
- The signal lives only in the phase, not the amplitude.
- Phase information
- Millimetric surface changes and coherence shifts between passes.
- Operational output
- Detect buried structures, tunnels, and route tracks.
Baltic · Strait of Hormuz
Moving target intelligence
- Problem
- Vessels go dark or spoof AIS, and noise defeats non-AI methods.
- Phase information
- Micro-Doppler and coherence patterns survive in the phase.
- Operational output
- Estimate motion and velocity of untracked vehicles and boats.
The same phase powers Europe's millimetric monitoring of dams, bridges and rail (Copernicus European Ground Motion Service). Context drawn from NATO critical-infrastructure operations, reported GNSS interference at major straits, and open SAR / InSAR literature.
Benchmarks — measured on SAR archive data
Measured gains before the mission model runs
+0 dB
Target contrast gap
TCR median: +7.32 dB vs +0.74 dB
The detector sees the target more clearly before classification.
0×
Class separation
Fisher: 0.413 vs 0.194
Fewer object classes get confused at the decision boundary.
+0pp
Speckle robustness
Accuracy in noise: 71.6% vs 53.2%
Works on raw SAR, not just cleaned imagery.
Measured on real SAR archive data, against a like-for-like classical baseline.
Our vision
Making raw radar directly usable by AI
Today the world’s radar archives are read as images, and most of the signal is discarded. Our goal is to own the path from raw signal to trusted intelligence, everywhere wave physics carries the answer. Radar first, physical signals beyond.
The company that makes raw radar waves usable by AI.
Defense Maritime Infrastructure
European, built for sovereign deployment.
Background & vision
Ruben Maarek
The technical foundation
Developed GPU-accelerated quantum simulation software at QbitSoft and complex algorithms at J.P. Morgan, specializing in processing high-dimensional mathematics on standard hardware.
Built the core engine
Independently coded and validated PerspeQtive's core quantum-inspired vision layer, achieving a proven +6.58 dB target contrast improvement on raw Level-1 SLC radar data.
Current focus & structure
Actively onboarding a CTO to deploy the AI models. Pre-incubation in Quantum Launchpad.
Founder & CEO
CentraleSupélec / NUS Singapore
Master in Quantum Machine Learning
Start here
See what your AI is missing on your archive radar data
What is hidden in your raw signals has value.
Request a data feasibility study
contact@perspeqtiveai.com
Start with a data study: we run our engine on your SAR archive and show you what the phase recovers.