Composer Interfaces

Multi-Vendor QPU Access

Composer Interfaces provides unified connectivity to quantum computing hardware from multiple vendors, abstracting the differences in APIs, transpilation requirements, and result formats behind a consistent interface.

Each supported backend is served by an independent worker service that handles the full lifecycle of circuit execution: transpilation to native gate sets, submission to the vendor API, result polling, and post-processing.

Supported Backends

Backend	Provider	Description
IBM Quantum	IBM	Access to IBM’s superconducting qubit processors via Qiskit Runtime
AWS Braket	Amazon	Access to IonQ, Rigetti, and other QPUs via Amazon Braket
IQM	IQM Finland	Access to IQM’s superconducting processors
LRZ	Leibniz Supercomputing Centre	Access to quantum resources hosted at LRZ

Architecture

Each vendor worker runs as an independent service that:

1. Receives jobs from the Composer Orchestrator via message queues
2. Transpiles circuits to the native gate set of the target device
3. Submits jobs to the vendor’s API
4. Polls for results and handles retries/timeouts
5. Returns results in a normalized format back to the orchestrator

  flowchart LR
    O["Composer Orchestrator"] --> W1["IBM Worker"]
    O --> W2["Braket Worker"]
    O --> W3["IQM Worker"]
    O --> W4["LRZ Worker"]
    W1 --> Q1{{"IBM QPU"}}
    W2 --> Q2{{"Braket QPU"}}
    W3 --> Q3{{"IQM QPU"}}
    W4 --> Q4{{"LRZ QPU"}}

Features

Unified Circuit Submission

All backends accept QASM circuits and handle vendor-specific transpilation internally. Users don’t need to worry about gate set differences or layout mapping.

Error Mitigation

Built-in support for measurement error mitigation (M3) across all backends, applied transparently during result post-processing.

Circuit Compression

Circuits are compressed during transmission to reduce network overhead, particularly important for large batch submissions.

Batch Processing

Multiple circuits can be submitted as a batch, with the worker managing parallel execution and result aggregation per backend.

Adding New Backends

New QPU providers can be integrated by implementing a worker service that conforms to the standard message interface:

1. Consume jobs from the orchestrator queue
2. Implement transpilation for the target gate set
3. Submit to the vendor API and poll for results
4. Return results in the normalized format

This modular architecture makes it straightforward to add support for new quantum hardware as it becomes available.