Experience quantum-enhanced AI search across satellite imagery, rare disease identification, traffic safety, and financial fraud detection datasets. Try our quantum models - no login required.
Advanced features for researchers and enterprises:
The Pinnacle QLDPC architecture reduces physical qubit requirements by 10x, enabling both RSA-2048 cryptanalysis and utility-scale quantum ML on the same networked QPU infrastructure.
RSA-2048 with < 100K physical qubits (10x fewer than surface codes). Modular QLDPC codes with Magic Engines for in-block distillation across 100-1,000 networked QPUs.
Squeezed-light FMCW extends cyclist detection from 20 m to 124 m (6.2x). 10x fewer chirps, 10x faster detection, 500 targets/sec throughput on unified TFLN photonic hardware.
44x rare-object recall via quantum ensemble optimization. 9.24x AUC/parameter efficiency with QRC. Fleet processing at 36x speed-up from 157K to 100M samples.
We are growing the team behind distributed quantum ML and AI platforms, hybrid quantum and classical delivery for defense, sensing, and industrial customers. United States roles: target August 2026. Switzerland / Europe roles: target September 2026.
View open rolesExperience the power of quantum computing for satellite imagery analysis. Search across 62 object classes using quantum-enhanced AI models - no login required.
This is a demonstration of quantum-enhanced search, a key capability of the quantum internet. The quantum models find rare objects that classical AI misses, showcasing quantum advantage in real-world applications.
Get full platform access to train quantum models, monitor real-time metrics, and deploy custom quantum-enhanced AI applications.
The Pinnacle architecture with QLDPC codes enables fault-tolerant quantum computing at 10x lower qubit cost — accelerating both cryptanalysis and real-world ML workloads on unified photonic hardware.
QLDPC codes in modular architectures encode multiple logical qubits per code block, unlike surface codes which encode only one. This fundamental advantage — combined with Magic Engines for in-block distillation and Clifford frame cleaning for parallel operations — yields an order-of-magnitude reduction in physical qubit requirements.
| Metric | Surface Code | QLDPC / Pinnacle |
|---|---|---|
| Logical qubits per block | 1 | k >> 1 (many per block) |
| Physical qubits per logical qubit | ~1,000 (at d=23) | ~100-200 (10x lower) |
| RSA-2048 physical qubits | ~1,000,000 | < 100,000 (10x fewer) |
| Fermi-Hubbard simulation | 940,000 qubits | 62,000 qubits (15x fewer) |
| Connectivity | Nearest-neighbor 2D grid | Quasi-local, modular/networked |
| Magic state distillation | Dedicated factories (high overhead) | Magic Engines (in-block) |
Pinnacle requirements: Physical error rate ≤ 10-3 | Code cycle: 1 μs | Reaction time: 10 μs | 100-1,000 networked QPUs × 100-1,000 qubits = 100K total for RSA-2048.
Ref: arXiv:2602.11457 (Webster et al., Feb 2026, Iceberg Quantum)
The same QLDPC fault-tolerant infrastructure that reduces RSA-2048 to < 100K physical qubits also enables utility-scale quantum ML. Distributed QPUs process LIDAR and radar data in real time, with networked quantum nodes handling compressed sensing, multi-sensor fusion, and rare-object classification at scales impossible for NISQ devices.
< 100K physical qubits via Pinnacle QLDPC — a 10x reduction over surface-code estimates (~1M qubits, Gidney 2025). Modular QPU networks with Magic Engines eliminate dedicated distillation factories.
Distributed QPUs for real-time FMCW LIDAR signal processing, compressed sensing at fault-tolerant scale, and multi-sensor LIDAR + Radar fusion across networked quantum nodes.
44x recall improvement for cyclists and pedestrians in LIDAR data using quantum ensemble optimization (QUBO). Scales from 157K to 100M samples with 36x speed-up vs XGBoost.
Each ML challenge in sensor perception maps directly to a quantum advantage enabled by the Pinnacle distributed architecture and Teraq's TFLN photonic platform.
| ML Problem | Quantum Technology | Key KPI | Improvement |
|---|---|---|---|
| Rare object recall (cyclists, pedestrians) | Quantum Ensemble (QUBO) | Recall @ 1% imbalance | 44x |
| Parameter efficiency | QRC (TFLN PIC) | AUC / 100K params | 9.24x |
| Fleet-scale processing (100M samples) | Quantum Ensemble (QUBO) | Throughput | 36x faster |
| Small-data accuracy (N=100 samples) | QRC (TFLN PIC) | Accuracy | +13% |
| Chipset FDTD optimization | QUBO optimizer (Dirac-3) | Optimization time | 35 days → < 1 sec |
| Safety warning time (@ 60 mph) | Squeezed-light FMCW + ML | React time | 0.74 s → 4.6 s (6x) |
Squeezed-light enhancement (5 dB, validated by DTU experimental data in arXiv 2502.07770) extends detection range by up to 6x for safety-critical low-RCS targets, while requiring 10x fewer chirps for equivalent detection probability.
| Target | RCS (m²) | Classical Range | Quantum Range | Improvement |
|---|---|---|---|---|
| Cyclist | 0.50 | 20 m | 124 m | 6.2x (+520%) |
| Pedestrian | 0.40 | 20 m | 111 m | 5.5x (+455%) |
| Vehicle | 25.0 | 20 m | 774 m | 38.7x (+3,768%) |
Teraq's TFLN photonic platform provides the physical interconnect layer for Pinnacle-class distributed quantum computing. The same chipset that generates squeezed light for quantum LIDAR also enables high-fidelity qubit links between networked QPUs.
As the number of exchanged qubits scales beyond 10, Teraq's efficient tomography methods (6x-40x sampling reduction, Richart et al., PRL 2010, Nature Comms 2024) unlock progressively more powerful sensing capabilities on the Pinnacle architecture.
| Qubit Scale | Capability Unlocked |
|---|---|
| 10-20 qubits | Hybrid Q-ViT for rare object detection (9x AUC/parameter) |
| 20-50 qubits | Quantum-enhanced FDTD chipset optimization |
| 50-100 qubits | Distributed FMCW coherent processing across sensor arrays |
| > 100 qubits | Quantum metrology network — SAR/LIDAR correlation at sub-cm precision |
Higher-dimensional encoding advantage: CGLMP encoding preserves ~2x bits/qubit under realistic noise. At 100 km inter-sensor distance, this yields a 1.47x information advantage over standard CHSH protocols, extending the effective range of distributed quantum sensing by 27-35%.
3 dB squeezing, hybrid quantum ML prototype. ST PIC100 at ~98% fidelity for hybrid QC-HPC (< 100 qubits).
6 dB squeezing, hybrid quantum ML production. TFLN at > 99% fidelity, > 100 qubits/link, demonstrated MHz rates.
6 dB qualified (AEC-Q100), fleet deployment, 1,000-10K units.
10 dB squeezing, QRC + hybrid Q-ViT fleet. 100-1,000 TFLN-networked QPUs → Pinnacle-class 100K qubits.
Try our live quantum LIDAR detection simulator or explore the quantum satellite search demo.
Quantum-enhanced AI models using Matrix Product State (MPS) quantum circuits with 8-qubit architectures. Currently deployed for satellite imagery analysis on the FMoW dataset, detecting rare objects across 62 classes with 92% recall.
Production models trained on FMoW satellite imagery dataset with Matrix Product State quantum circuits.
Matrix Product State Architecture
Flagship quantum-enhanced satellite detector with 80M parameters. Uses 8-qubit Matrix Product State circuits for quantum feature extraction, achieving 95.4% recall on rare objects - 16% higher than classical baselines (F1: 0.60 vs 0.23, AUC: 0.987 vs 0.875).
ResNet + Quantum Layers
Hybrid architecture combining ResNet-50 classical backbone with quantum enhancement layers. Achieves 95.4% recall with 0.60 F1 score and 0.987 AUC. Optimized for edge deployment with lower latency while maintaining high quantum performance.
NASA/IBM Foundation Model (ViT)
Vision Transformer foundation model from NASA and IBM, pre-trained on Harmonized Landsat Sentinel (HLS) data. Classical baseline achieving 79.3% recall (F1: 0.23, AUC: 0.875), demonstrating 16% quantum advantage gap in rare object detection.
Experience quantum-enhanced AI search across satellite imagery, rare disease identification, traffic safety, and financial fraud detection datasets.
Already have access? Login here
Fill out the form below to request access to the TERAQ quantum computing platform. We'll review your application and get back to you within 24 hours.
Get in touch with our quantum computing experts for support, partnerships, or technical inquiries.