en.Wedoany.com Reported - Quantum cloud platform qBraid announced a series of infrastructure expansions and algorithmic breakthroughs aimed at strengthening its hybrid quantum-classical development pipeline. These updates make qBraid a remote cloud target within the NVIDIA CUDA-Q framework, expand the on-demand GPU hardware fleet for qBraid Lab, and deploy Google Cloud's AlphaEvolve automated coding agent to address resource bottlenecks in fault-tolerant quantum chemistry simulations.
Through its integration as a remote cloud target within NVIDIA CUDA-Q, developers can use the native nvq++ compiler toolchain to directly compile quantum kernels and dispatch them to physical hardware supported by qBraid. This architecture enables users to target hardware backends from vendors such as Rigetti, IonQ, IQM, and QuEra by adjusting the machine flag in execution statements using a single qBraid API key. The pipeline includes access to qBraid's free quantum intermediate representation state vector simulator, supporting asynchronous submission and disk future persistence for workloads up to 30 qubits and 2000 shots.
To support intensive hybrid workloads such as tensor network simulations, variational optimization, and neural network error correction decoding, qBraid Lab has expanded its infrastructure to provide on-demand access to over 20 GPU instance types. The pay-as-you-go fleet, orchestrated by qBraid CTO Ryan Hill, eliminates the friction of reserved capacity, allowing users to launch configurations directly from a browser-based JupyterLab or VS Code environment. Available compute tiers span multiple hardware generations, including: Blackwell architecture NVIDIA B200, Hopper architecture NVIDIA H200, NVIDIA H100, and NVIDIA GH200 Grace Hopper Superchip, Ampere and Ada Lovelace architecture NVIDIA A100, NVIDIA L4, NVIDIA L40S, RTX 4090, RTX 5090, and RTX 6000 Ada. These instance profiles support native execution of dedicated quantum calibration models, such as the NVIDIA Ising open AI series, which are pre-configured to run with the CUDA-Q compilation stack.
Targeting the foundational mathematical layer of quantum chemistry, qBraid's research team—including Dr. Kenny Haydrite, James Brown, and Tarini Hadika—collaborated with Google Cloud's AlphaEvolve early access program to optimize fermion-to-qubit encoding. Due to the exponential search space, with over 10^50 possible configurations for an 8-orbital molecule, converting molecular electronic structure into qubit operators presents a significant design challenge. Leveraging the Gemini model within an evolutionary loop, the AlphaEvolve agent iteratively modified seed Python structures based on qBraid's proprietary Generalized Superfast Encoding family, evaluating approximately 1500 program variants against a rigorous and unbreakable exact validator scoreboard. The resulting AI-generated encoding rules successfully bypassed traditional manual design constraints, achieving an exact quantum error correction code distance of 5 on dense molecular Hamiltonians, whereas the previous best human-engineered solution only reached distance 3. When validated against held-out chemical systems not encountered during model training, such as beryllium hydride and water, the generated code maintained distance-5 protection. The newly discovered structures reduced logical error rates by a factor of 3.4 to 7.9 under exact decoding, while requiring 4.2 to 5.0 times fewer data qubits than standard fault-tolerant compilation paths, thereby reducing the physical hardware overhead needed for deep molecular simulations.
This article is compiled by Wedoany. All AI citations must indicate the source as "Wedoany". If there is any infringement or other issues, please notify us promptly, and we will modify or delete it accordingly. Email: news@wedoany.com
