- Unitary Quantum launched UnitarySpark, a desktop workstation integrating quantum simulation, GPU acceleration, AI-agent interaction and real-machine interconnection for hybrid classical-quantum computing.
- UnitaryLab 2.5 entered public beta with a natural-language AI agent that converts scientific problems into executable quantum workflows, reducing manual coding steps by roughly 75 percent according to an NVIDIA case study.
- The platform is built on Schrodingerization, an original algorithm framework from Shanghai Jiao Tong University that maps continuous differential equations onto unitary quantum evolution.
- All inference, storage and execution occur locally on the workstation, addressing data-sensitivity requirements in finance, drug design and other regulated fields.
Desktop Hardware Meets AI-Agent Interface
At a September 20 launch event in Shanghai, Unitary Quantum introduced UnitarySpark, described as a first-generation heterogeneous computing base product. The workstation integrates quantum simulation, GPU-accelerated classical computing, an AI-agent layer and a unified Circuit API that connects simulators to real quantum processors without code changes.
Alongside the hardware, the company opened public beta testing for UnitaryLab 2.5, a platform where users describe scientific problems in natural language. An embedded AI agent then translates those descriptions into mathematical formulations, selects algorithms, schedules compute resources and returns interpreted results. According to an NVIDIA success case published in September 2026, this agent workflow compressed approximately twelve manual coding steps into three natural-language prompts, cutting interaction frequency by about 75 percent.
Co-founder Professor Zhang Lei said the agent serves as a dispatch hub between researchers and underlying compute, rather than a simple chat interface, aiming to lower the barrier to using quantum computing for scientific research.
Local Data and Real-Machine Interconnection
CTO Dr. Hu Junpeng described UnitarySpark as a desktop quantum-computing workstation rather than a standalone universal quantum computer. All reasoning, algorithm execution and data storage happen on the local machine, so experimental parameters and model configurations never leave the premises, which is a requirement for compliance-sensitive domains such as financial modeling and pharmaceutical research.
The workstation also provides a pathway from simulation to physical hardware. Through its unified Circuit API, algorithms validated in simulation can be transferred directly to connected real quantum processors. Unitary Quantum announced strategic partnerships at the event with Shanghai-based quantum hardware firms including Guodun Quantum, Bose Yuanpei Quantum and Zhongqi Wuliang Quantum to validate this simulation-to-real-machine pipeline.
Algorithm Foundation and Industry Validation
The platform rests on Schrodingerization, an original algorithm framework developed by the Shanghai Jiao Tong University quantum scientific computing team. The method applies a mathematical transformation that converts non-unitary differential equation problems into unitary evolution forms that quantum systems can process directly, offering potential algebraic or exponential speedups for high-dimensional, multi-scale scenarios. The algorithm was included in the National Natural Science Foundation of China 14th Five-Year Plan outstanding results compilation as the sole representative mathematics achievement in 2024.
UnitaryLab implementation on NVIDIA DGX Spark was selected as a success case on the NVIDIA website and included in the first batch of the AI for Industry capability list, earning recognition as a 2026 exemplary domestic industrial simulation software case. At the launch event, additional industry partners signed cooperation agreements covering finance, meteorology, biomedicine, artificial intelligence and quantum education.
Sources & context
Go to the original material. Company claims remain attributed to their sources.
01
