ML Ops Engineer, Quantum Machine Learning

About the role

We are hiring an ML Ops Engineer into Quantum Machine Learning, the team that runs a production QML reservoir on the Watermelon chip: quantum feature generation that improves deep learning on customer problems, plus the models and pipelines around it. The scientists tune the physics, the devices and the models. You will build what the science depends on: the data path in, the training and evaluation pipelines, and the metrics that feed the roadmap.

The feature extractor is a quantum device with a queue and a calibrated state, so every training run has to record which chip and which calibration produced its features. Benchmarking devices and models against each other matters to our R&D work and to the results improving our device design.

Customer engagements drive the work. A dataset arrives, often needing data engineering and usually under NDA, and the team has to get from raw tables to a trained model with a defensible comparison against whatever the customer uses today. Doing that once is a research project. This role is doing it repeatedly, for several customers at a time, without the pipelines rotting between engagements.

Based at our Sydney facility, you will work across the science, the platform and the customer-facing teams. This is a role for someone who can move research code into production without slowing the research down, and who wants to build benchmarks that stand up to scrutiny from a customer's own data scientists.

Role responsibilities

• Own the data path from a customer dataset landing to features ready for training: ingestion, validation, cleaning, splitting, versioning and access control
• Build and run the training and evaluation pipelines that take quantum features from the Watermelon chip through to a trained, measured model
• Make experiments reproducible: track datasets, hyperparameters, code versions, chip identity and calibration state alongside every result
• Build the benchmarking harness that compares quantum features against classical deep learning baselines on equal terms, and keep it valid as models and datasets change
• Automate the path onto hardware: batching jobs against device queues, handling device unavailability and re-runs, and reconciling feature sets generated across different chip states
• Deploy models into customer-facing serving, and monitor them for accuracy, latency and data drift
• Build the pipeline scaffolding that lets a new customer engagement start from a working template instead of a blank notebook
• Instrument the pipelines so the team can see cost, throughput, device utilisation and where a run spent its time
• Work with Integration & Access on the QML APIs and SDKs, feeding back what the pipelines need from the platform
• Handle customer data under its commercial and privacy obligations, including segregation, retention and audit
• Balance model explainability, complexity and function when reporting results, considering more than model performance alone
• Document pipelines, data lineage and operational procedures so someone who did not run a result can reproduce it

Your experience

Essential

• 4+ years building and operating machine learning pipelines or data platforms in production
• Strong Python, including the scientific and ML stack (NumPy, pandas, PyTorch or JAX, scikit-learn)
• Data engineering across batch and streaming workloads: pipeline orchestration (Airflow, Dagster, Prefect or equivalent), schemas, data validation and versioning
• Practical MLOps: experiment tracking, model registries, reproducible training, CI/CD for models, and monitoring after deployment
• Sound grasp of ML methodology, particularly evaluation: train/test hygiene, cross-validation, leakage, and how to build a benchmark that stands up to scrutiny from a customer's own data scientists
• Containers, and running workloads on infrastructure shared with other teams
• Git workflow, code review and testing, applied to research code as well as production code
• Working with data that carries commercial or regulatory constraints
• Working directly with commercial customers on their data and their problems
• Clear technical writing, and the ability to explain a result and its caveats to people who are not ML specialists

Nice to have

• MLflow, or a comparable experiment tracking and model registry stack (Weights & Biases, Neptune)
• Feature stores, or any system where derived features need provenance back to their inputs
• Time series or signal data, which is where much of our customer work sits
• Reservoir computing, kernel methods, or other approaches that treat a fixed nonlinear map as a feature generator
• Quantum computing exposure, including any of Qiskit, PennyLane, Cirq or similar (no physics degree required)
• Hardware-in-the-loop pipelines, or scheduling work against instruments and devices with queues and downtime
• HPC and batch schedulers such as Slurm, and GPU cluster work
• Model serving at low latency (Triton, Ray Serve, TorchServe or equivalent)
• Data engineering at scale (Spark, Dask, DuckDB, Parquet and columnar formats)
• Cost and performance tuning across compute, storage and accelerator time
• Rust or Go alongside Python