Quantum Pipeline

Context

Since I started my engineering degree, quantum computing has been one of my interests. To pursue this - I started building something that combined study with a modern tech stack - a project that could be a subject of an engineering thesis and serve as a portfolio piece.

Approach

Simulation

The starting point was a simulation module for the Variational Quantum Eigensolver (VQE). It seemed appropriate for study: both classical and quantum components, controlled experiments and possibility to utilise IBM Quantum Experience free credits.

GPU acceleration

Next came CUDA-based GPU acceleration through Qiskit Aer, which had already been tested at that stage. Getting the GPU Docker image running on a dedicated instance proved more challenging than expected - NVIDIA drivers, and some compatibility issues with the precompiled version of Aer from pip.

Having a separate machine for Aer's compilation and testing made a difference. In case something didn't go as planned - it was a problem for another day.

Integration with Kafka

A simulation module on its own is interesting but not particularly useful. The data it produces should be somehow forwarded to storage and some sort of analytics. Apache Kafka and Spark were a natural fit - something worth learning independently of this project.

The remaining work leaned more toward infrastructure: connecting things together so the data the simulation produces can be utilised. The specific challenges there might warrant separate articles.

Outcome

After a long break spent writing the thesis, development resumed. Alongside a master's degree, the aim is to bring it closer to a proper end-to-end system.

Recent work focused on increasing test coverage, writing documentation, addressing problems in the simulation module that the thesis revealed, and cleaning up the CI/CD pipeline - making it more reliable and less redundant.

GitHub · Codeberg (mirror) · PyPI · DockerHub · Documentation

S.D.G