Qedma integrates QESEM with NVIDIA CUDA-Q platform

Key Takeaways

• Qedma’s QESEM error‑mitigation engine is now natively supported inside NVIDIA’s CUDA‑Q quantum‑programming stack.
• The integration currently targets Quantinuum trapped‑ion processors; extensions to IBM, Rigetti, and other platforms are slated for 2025.
• QESEM’s patented “Quantum Error Subspace Expansion” (QES) reduces observable error rates by ≈ 60‑70 % on benchmark circuits, enabling deeper algorithms without additional qubits.
• CUDA‑Q developers can call QESEM via the same C++/Python API they already use—no code refactor required.
• Early adopters include Fortune‑500 manufacturers, top‑ranked universities, and national labs, all reporting up to 3× improvement in result fidelity.

Introduction: Bridging Two Quantum Leaders

Qedma Quantum Computing, a specialist in quantum‑error mitigation, announced that its QESEM (Quantum Error Subspace Expansion Mitigation) software is now integrated with NVIDIA CUDA‑Q, the GPU‑accelerated quantum programming platform. This partnership merges Qedma’s noise‑reduction algorithms with CUDA‑Q’s hardware‑agnostic compiler and runtime, giving developers a single environment to write, simulate, and execute quantum workloads with built‑in error mitigation.

What Is QESEM?

QESEM is a patented error‑mitigation technique that builds a low‑dimensional subspace around the noisy quantum state and solves a classical least‑squares problem to extrapolate the ideal expectation values. Key specifications include:

Data derived from Qedma internal benchmarks (Q‑2024) and independent academic studies.

How CUDA‑Q Benefits From QESEM

Seamless API Integration

CUDA‑Q users can invoke QESEM through a single function call:

import cudaq
result = cudaq.sample(my_circuit, error_mitigation="qe_sem")

No changes to kernel definitions, memory management, or compilation flags are required. The mitigation step runs on the same GPU that hosts the CUDA‑Q kernel, preserving low‑latency data flow.

Hardware‑Agnostic Execution

While the initial release is validated on Quantinuum H2 trapped‑ion devices (127 µs gate time, 99.9 % single‑qubit fidelity), the abstraction layer abstracts hardware specifics. Future updates will auto‑detect IBM Falcon or Rigetti Aspen back‑ends and apply the appropriate QESEM calibration data.

Performance Footprint

• GPU Memory Impact: < 50 MiB per mitigation job (negligible for modern RTX 4090‑class cards).
• Latency Penalty: Average added wall‑time of 3‑5 ms per circuit, dominated by the classical linear‑solver step.
• Throughput: Benchmarks on a 24‑core Xeon 8360Y + RTX 4090 system show a 2.2× speed‑up over a pure CPU‑only mitigation pipeline.

Real‑World Adoption

• Automotive Supplier (2024 Q3): Integrated QESEM‑CUDA‑Q into a quantum‑inspired routing optimizer, achieving a 2.8× reduction in solution variance.
• MIT Quantum Lab: Used the combined stack to run a 20‑qubit variational quantum eigensolver (VQE) on Quantinuum hardware, reporting a 3.1× increase in energy‑estimate accuracy versus raw measurements.
• U.S. Department of Energy (DOE) Lab: Deployed QESEM within CUDA‑Q for quantum chemistry simulations, cutting required sampling shots from 10⁶ to 3 × 10⁵ while maintaining chemical‑accuracy thresholds (< 1 kcal/mol).

Future Roadmap

• 2025 Q1: Full support for IBM Quantum System One (27‑qubit) and Rigetti Aspen‑10.
• 2025 Q3: Introduction of “adaptive QESEM,” which dynamically adjusts subspace dimension based on real‑time noise diagnostics.
• 2026: Co‑development of a CUDA‑Q‑native tensor‑network simulator that leverages QESEM’s error‑model to produce more realistic noisy‑state predictions.

Bottom Line

The Qedma‑CUDA‑Q integration delivers a pragmatic path to higher‑fidelity quantum computing today, sidestepping the wait for fault‑tolerant hardware. By embedding QESEM’s advanced subspace‑expansion mitigation directly into NVIDIA’s GPU‑accelerated development stack, users gain up to a 70 % reduction in noise‑induced error with only a modest millisecond overhead. Early adopters across industry, academia, and government are already seeing tangible gains in algorithmic depth and result reliability. As the partnership expands to additional quantum processors, QESEM‑CUDA‑Q is poised to become a standard component of any serious quantum‑software workflow.