Quantum Machine Learning Researcher Job Description

A Quantum Machine Learning Researcher works where quantum computing and machine learning overlap. One direction asks whether a quantum processor can learn from data better than a classical computer: the researcher designs parameterized quantum circuits, chooses how classical data gets encoded into qubit states, and trains the result as a model. The other direction runs the opposite way, using machine learning to solve problems inside quantum computing itself, such as reading out noisy measurements, characterizing device errors, or decoding error-correction syndromes. Some researchers work in both.

The tooling is specific. The open-source Qiskit ecosystem includes Qiskit Machine Learning, which the project describes as providing "fundamental computational building blocks, such as Quantum Kernels and Quantum Neural Networks, used in various applications including classification and regression." Other groups work in PennyLane, Cirq or GPU-accelerated simulators. A typical week mixes classical simulation on a workstation or cluster, a smaller number of runs on a cloud quantum processor, and a lot of analysis about why a model did or did not train.

That last point defines the job. Many quantum models face the barren plateau problem, where the loss landscape flattens as systems grow, gradients shrink toward zero and training stalls. A researcher spends real time proving or measuring whether a proposed architecture avoids it, then checking whether a classical method can match the result anyway. A claimed speedup that survives a strong classical baseline is the deliverable; one that does not is still a publishable negative result.

One proven advantage result concerns quantum data rather than spreadsheets or images. A 2022 paper in Science reported, "We proved that quantum machines could learn from exponentially fewer experiments than the number required by conventional experiments," for tasks such as predicting properties of physical systems. Results of that kind apply where the input is already quantum, such as data produced by physical systems and experiments.

The researcher rarely works alone. At NASA's Quantum Application and Information Laboratory (QuAIL), the team tracks and contributes to algorithms for "optimization, machine learning, simulation of condensed-matter, high-energy physics, chemistry and materials," and NASA notes that while the group is a theory and numerical team, "a key component of this work is close collaboration with quantum hardware groups." In practice that means translating qubit connectivity, gate error rates and readout limits into models the hardware can actually run, and feeding algorithm needs back to device engineers.

This is a different job from classical machine learning research. A classical ML scientist assumes the hardware works and asks how to train a better model on GPUs at scale. A Quantum Machine Learning Researcher cannot assume the hardware helps at all: the open question is whether a quantum resource adds anything, and much of the work is the careful bookkeeping that answers it honestly. Output takes the form of papers, preprints, open-source code, resource estimates and internal technical reports for the hardware roadmap.

As a rough guide, a Quantum Machine Learning Researcher holds a PhD in physics, computer science, applied mathematics or electrical engineering, often followed by one or two postdoctoral appointments. The research is closer to what O*NET describes for computer and information research scientists, who "conduct research into fundamental computer and information science as theorists, designers, or inventors," than to an applied engineering role. BLS gives a master's degree as the usual entry-level credential for that broader occupation, though it adds that some employers favor candidates holding a doctorate. As a rough guide, a doctorate or an equivalent publication record is the practical bar for a research scientist title in quantum machine learning. Engineers with a master's degree can enter through research software or quantum application roles and build toward research work.

The core knowledge comes in three layers. First, quantum information: qubits, gates, measurement, density matrices, noise channels and the basics of error correction. Second, statistical learning theory and modern machine learning: kernels, generalization bounds, optimization, neural network architectures and training diagnostics. Third, the linear algebra and probability that connect them, since most results in the field are stated as bounds on sample complexity, expressivity or gradient variance.

Practical skills matter as much as theory. Expect to write production-quality Python, work in at least one quantum SDK such as Qiskit, PennyLane or Cirq, and use PyTorch or JAX for the classical side. Experience with tensor-network or state-vector simulators, high-performance computing schedulers and GPU programming helps, because experiments run in simulation before they touch hardware. A researcher who can run error mitigation on a real device and explain the calibration data behind a result is useful to hardware teams.

There is no license, and no certification stands in for a research record. A strong application rests on first-author papers in physics or machine-learning venues, open-source contributions, and evidence that the candidate has benchmarked quantum models honestly against classical ones. A careful negative result is a credible thesis chapter; an unverified speedup claim is a liability.

Typical paths in: a physics PhD with a machine learning thesis chapter, a computer science PhD in learning theory that picked up quantum information, or a chemistry or materials PhD that used quantum simulation. National laboratory postdocs, university quantum centers and industry residency programs serve as bridges into a staff researcher role. As a rough guide, staff research scientist positions in industry follow two to five years of postdoctoral or equivalent research experience.

Soft skills are concrete here: writing clearly for referees who come from a different field, scoping projects with chemists or physicists who own the application, and telling a hardware team plainly when a model will not fit their device.

BLS does not track quantum machine learning as its own occupation, so the closest federal projection covers all computer and information research scientists. The Occupational Outlook Handbook projects that occupation to "grow 22 percent from 2025 to 2035, much faster than the average for all occupations," and about 2,900 openings a year on average. The same handbook lists the federal government, excluding postal service, at 31% of employment, computer systems design and related services at 16%, and research and development in the physical, engineering, and life sciences at 14%.

The field itself is young. Argonne National Laboratory describes quantum machine learning as "an emerging, exciting field for addressing previously untenable computational tasks," and its Quantum Institute lists a project on quantum machine learning based on multimode superconducting piezomechanics. Readers weighing a career move should take the word emerging literally: the research questions are open, and so is the timeline for commercial use.

Federal policy moved in 2026. Executive Order 14413, signed June 22, 2026, directed an update to the National Quantum Strategy and established the Quantum Computer for Application Development and Discovery Science (QC-ADDS) effort, which "shall pursue development of a quantum computer at a scale intended to initiate the era of quantum-enabled scientific discovery." The order also includes a section on expanding and retaining the quantum workforce. On Sept. 17, 2026, the National Quantum Coordination Office reported that the Department of Energy had launched the Quantum Genesis Q Competition, "with up to $215 million in planned funding," and NSF announced eight Quantum Leap Challenge Institutes on Aug. 26, 2026. The order and the DOE competition target quantum hardware and scientific applications broadly; neither names quantum machine learning.

The AI side of the job has its own research track. A 2024 Nature paper described AlphaQubit, a recurrent, transformer-based neural network that "learns to decode the surface code," and reported that it outperformed other state-of-the-art decoders on real-world data from Google's Sycamore processor for distance-3 and distance-5 surface codes. Work of this kind, machine learning in service of fault-tolerant hardware, gives researchers a path that does not depend on proving a quantum speedup for classical data.

Career paths split three ways. A researcher can stay on the individual contributor track, moving from research scientist to senior and principal scientist with a larger say over which problems a lab pursues. A second path leads into research management, running a quantum algorithms or applications group. A third moves toward applied work: resource estimation, customer application pilots with chemistry or finance teams, or quantum software products. Academic routes remain open as well, through faculty positions at universities with quantum centers. Researchers who keep strong classical machine learning skills keep options in conventional AI research if quantum timelines slip, which is a practical hedge worth planning for from the start.

Dear Hiring Committee,

I am applying for the Quantum Machine Learning Researcher position on your quantum algorithms team. I am completing a postdoctoral appointment in a university quantum information group, where my work sits between variational quantum algorithms and statistical learning theory.

My doctoral research in physics studied quantum kernel methods for classifying phases of matter from measurement data. The most useful thing I learned was how to lose an argument with a classical baseline. Two of my early models looked promising on small simulated systems, and both were matched by a classical kernel once I tuned it properly. I wrote that up as a short paper, and the analysis behind it became the benchmarking harness our group now uses for every new proposal.

Since then I have focused on problems where the input is quantum to begin with. I built a pipeline that trains shallow circuits on data produced by a partner group's superconducting device, with error mitigation and calibration records stored next to every result so that a referee or a hardware engineer can reproduce a figure from raw shots. I wrote the code in Qiskit and JAX and maintain it as an open-source package with documentation and tests.

I also work on the other direction of the field. With a colleague in the hardware group, I trained a small neural network to flag drifting qubits from routine calibration data, which saved the team manual checks before overnight runs.

Your group's focus on algorithm and hardware co-design is the reason I am applying. I communicate well with device engineers, I am comfortable telling a collaborator when an idea will not fit the hardware, and I care about results that hold up.

Thank you for your consideration. I would welcome the chance to discuss how I could contribute.

Sincerely, Jordan Avery