Oracle and Classiq technologies
Oracle and Classiq have introduced a successful proof of concept (PoC) that combines quantum expert AI agents with high-performance conventional computing infrastructure, which represents a major advancement for the quantum software engineering. This partnership shows an end-to-end process that uses large-scale simulations on Oracle Cloud Infrastructure (OCI) to develop complicated quantum applications from straightforward natural-language signs.
You can also read Classiq Technologies Launches AI Agent for Quantum Software
Closing the Execution and Intent Divide
The high level of expertise required to develop efficient algorithms and the massive processing power required to evaluate them at scale are usually the two primary barriers to the development of quantum software. By linking Classiq’s AI-powered synthesis engine with Oracle’s powerful GPU nodes, the collaboration between the two companies addresses both issues.
Start with a natural-language purpose, construct a structured development environment, create a non-trivial quantum circuit, then use OCI to simulate on NVIDIA GPUs to produce a practical process. With very little human parameter adjustment, the team was able to create a workable solution for a challenging portfolio optimization problem by using this automated process.
You can also read Classiq and NVIDIA CUDA-Q Raise Hybrid Quantum Applications
High-Performance Simulation’s Crucial Role
Before switching to real quantum hardware, teams may test models, troubleshoot circuits, and evaluate algorithm options using simulation, which is an essential link in the quantum development lifecycle. However, as the number of qubits rises, these simulations’ processing requirements increase gradually.
Although Classiq’s basic flow typically allows simulations of up to 29 qubits, this proof of concept extended the boundaries to 36 qubits. Using eight A100 GPUs operating in parallel, Classiq-generated circuits were routed to an NVIDIA DGX A100 node hosted on OCI to reach this scalability.
Such a task requires enormous amounts of memory. There are around 68.7 billion complex amplitudes in a 36-qubit state vector. This state vector alone takes about 512 GiB of GPU RAM to store at single precision; this amount does not account for the extra overhead needed by the simulator runtime. This requires the implementation of a distributed, multi-GPU architecture, like the one provided by Oracle, to maintain the necessary aggregate memory.
You can also read Classiq Quantum Simplifies Complex Chaos Modeling with Qmod
From AI Prompt to Prototype in 15 Minutes
The demonstration’s rapid evolution was among its most noticeable features. Classiq’s expert AI agent, which is educated by a vast library of algorithms and business applications, was given a natural-language signals to start the process.
The request requires a Markowitz-model portfolio optimization software for 12 assets, with particular limits on financial constraints, risk-aversion variables, and asset allocation sections. The AI agent created an extensive Jupyter notebook with the problem setup, a hybrid classical-quantum process, and tools for result analysis in less than fifteen minutes. Fast prototyping allows developers to focus on high-level integration and model selection by significantly reducing the amount of time spent on “boilerplate” code.
You can also read Classiq’s Qmod Quantum Conditionals in Grover’s Algorithm
A 36-Qubit Financial Optimization Workflow
The particular application modeled twelve correlated assets from different industries and focused on a discrete version of a Markowitz-style optimization. The budget for the whole portfolio may be set at 15 sections, with each asset being allotted between zero and seven. As a result, there were 68.7 billion potential combinations in the search area.
The procedure addressed this using the Quantum Approximate Optimization Algorithm (QAOA), a hybrid method in which a quantum circuit samples solutions and a classical optimizer—COBYLA updates circuit parameters. The encoding approach, which required three qubits per asset to represent the eight different allocation levels, directly led to the 36-qubit circuit size. Because of the platform’s high-level algorithmic structures, Classiq’s synthesis engine produced a circuit with a depth of 730 that was nevertheless understandable and flexible.
Smooth Integration with OCI
To send Classiq circuits directly to the OCI host via SSH, the technical execution depended on a lightweight adaption layer. The procedure entailed:
- Binding QAOA parameters and exporting the circuit.
- Uploading auxiliary scripts and OpenQASM files to the Oracle environment.
- Starting a Docker container with the NVIDIA cuQuantum Appliance.
- Distributing the workload among all eight GPUs using mpirun.
The notebook queried the OCI directory to get result artifacts over the five-hour simulation, ensuring a smooth developer experience where Oracle’s infrastructure took care of the hard work and the notebook remained the main interface.
You can also read Classiq Quantum computing with AMD, Comcast for Internet
Analysis of Results
15 outer optimizer iterations and 16,384 shots per sampling call were used in the hybrid simulation. A conventional benchmark of 200,000 random viable allocations was used to compare the final quantum-sampled allocation.
With a mere 4.63% objective-function gap in comparison to the classical reference, the quantum outcome was determined to be a robust allocation. More significantly, the PoC demonstrated that businesses can now use existing infrastructure to scale their quantum operations and get significant results, guaranteeing model portability as quantum technology advances.
Quantum Software Engineering’s Future
This partnership is a game-changer for the sector. Oracle and Classiq have significantly lowered entry barriers by fusing scalable simulation infrastructure with AI-assisted development. Researchers and developers from a variety of industries, including cybersecurity, healthcare, aerospace, and defense, may now use this “enterprise-grade workflow” to expedite their optimization efforts.
