The public image of a quantum computing company is a room of theoretical physicists. Look at the job board of an actual one and you will mostly find engineers.
Who actually works in a quantum computing company
Building a quantum computer is an engineering project with a physics core, and the ratio favours engineering heavily.
Cryogenic engineers. Superconducting machines operate at temperatures a few thousandths of a degree above absolute zero. Achieving that, maintaining it, and getting hundreds of control lines into a cold space without carrying heat down them is a substantial mechanical and thermal engineering problem.
Microwave and RF engineers. Superconducting qubits are controlled with precisely shaped microwave pulses. Signal generation, timing, filtering, attenuation and isolation are ordinary RF engineering performed to unusual tolerances.
Control systems and electronics engineers. Fast, synchronised, low noise electronics delivering thousands of coordinated pulses, increasingly with feedback inside a single computation.
Fabrication and process engineers. Superconducting qubits are made with lithography in cleanrooms, and the skills overlap heavily with semiconductor processing.
Software engineers. Compilers, schedulers, calibration automation, cloud access layers, simulators. Much of this is ordinary systems software with an unusual target.
Systems and integration engineers. Making a machine that is a laboratory experiment behave like a product.
Almost every one of these roles is a recognisable engineering discipline applied to an unusual object.
Physicists matter enormously in the core of the field. They are not where most of the hiring is, and treating a physics doctorate as the entry requirement excludes most of the people the industry actually needs.
What the direction covers
The scope: quantum mechanics fundamentals, qubit technologies, quantum algorithms, quantum software and the engineering of quantum hardware.
Four areas.
Fundamentals. Superposition, entanglement, measurement and decoherence, at the level required to reason rather than to derive.
Hardware platforms. Superconducting circuits, trapped ions, neutral atoms, photonic approaches and semiconductor spin qubits, each with different strengths and failure modes.
Algorithms. What is known to give advantage, under what conditions, and with what resource requirements.
Engineering. Control, cryogenics, calibration, error correction and the systems work that turns any of it into a machine.
Error correction, and why the qubit count is not the number
The number quoted in announcements is usually physical qubits. It is close to meaningless on its own.
Qubits lose their state. They interact with their environment, they drift, and operations on them are imperfect. The useful computation time before the information degrades is finite and short, which bounds how long an algorithm can run.
Quantum error correction addresses this by encoding one reliable logical qubit across many physical qubits, with continuous measurement detecting and correcting errors without destroying the computation. It works in principle and has been demonstrated at small scale.
The cost is the point. The ratio of physical qubits to one useful logical qubit is large, and it depends on the error rate of the underlying hardware: better qubits need fewer of them. Estimates for the physical qubit counts required to run the algorithms people actually want are far above what exists today, which is why the gap between current machines and commercially transformative ones is a matter of orders of magnitude rather than iterations.
Two honest consequences.
Qubit quality matters more than quantity. Error rate, coherence time and connectivity determine how many physical qubits an error corrected system needs. A machine with more, worse qubits can be further from useful than one with fewer, better ones.
Nobody can responsibly give a date. The remaining problems are hard engineering problems with genuine uncertainty in them. Predictions in this field have a poor track record in both directions, and a confident timeline should be treated as a commercial statement rather than a technical one.
Where this sits in the domain
Quantum computing is the eighth of nine directions in Astra Trainer's semiconductors, electronics and quantum domain. It draws directly on RF, wireless and telecommunications for microwave control, on semiconductor manufacturing for qubit fabrication, on photonics for photonic and trapped ion approaches, and on electronics engineering for the control stack.
It pairs closely with quantum communication, sensing and security, which is where the nearest term practical work sits. Partners usually scope quantum computing alongside one of the engineering directions, because that is where their existing people can actually contribute. You can see the nine directions here.
What quantum computers are and are not good at
The most useful thing training can do in this direction is prevent expensive misunderstanding.
A quantum computer is not a faster computer. It does not speed up general computation, databases, machine learning as a category, or anything simply because the problem is large.
The known advantages are specific.
Simulating quantum systems. The original motivation and the most credible application. Molecular and materials behaviour is quantum mechanical, and classical simulation of it becomes intractable quickly. Chemistry and materials discovery are the areas where advantage is most plausible.
Certain algebraic problems. Factoring large numbers and computing discrete logarithms, which is exactly why the cryptographic implications are taken seriously.
Some optimisation and sampling problems, where the picture is genuinely mixed. Advantage over the best classical methods has been harder to demonstrate than early claims suggested, and several proposed advantages have been eroded by improved classical algorithms.
What follows for planning is unglamorous and correct. If an organisation's hard problems are quantum chemistry or materials, there is a reason to build understanding now. If they are logistics or scheduling, the case is weaker and the classical alternatives keep improving. If they involve long lived confidential data, the relevant direction is the cryptographic one, and that is not speculative.
What an organisation can reasonably do now
Four actions that are defensible without assuming any particular timeline.
Build literacy in the people who would evaluate claims. The most common failure is not missing an opportunity. It is being unable to assess a vendor claim, which leads either to overcommitment or to dismissing something real.
Identify whether your problems are in the plausible set. A short exercise with someone who knows the algorithms saves years of vague interest.
Develop the engineering skills that are useful regardless. Cryogenics, microwave engineering, precision control and low noise electronics are valuable in many industries. Training people in them is not a bet on quantum computing.
Treat cryptography as the exception. It is the one area where inaction has a cost today, for the reason covered in the neighbouring direction: encrypted data captured now can be stored and decrypted later.
The roles, named
Quantum hardware engineers. Device design, fabrication, characterisation.
Cryogenic engineers. Dilution refrigeration, thermal design, wiring.
Microwave and control electronics engineers.
Quantum software engineers. Compilers, tooling, simulators, cloud platforms.
Quantum algorithm researchers. The smallest group and the most visible.
Calibration and characterisation engineers, who keep machines working day to day.
Error correction engineers, bridging theory and implementation.
Applications specialists, translating between domain problems and quantum formulations.
Systems and integration engineers, turning experiments into products.
Who can be trained into it
RF and microwave engineers. The most direct conversion in the whole direction. Qubit control is microwave engineering to demanding tolerances, and the required physics is learnable by someone who already understands signals.
Cryogenics and vacuum engineers. Directly applicable, from physics laboratories, industrial gas, space testing and superconducting magnet work.
Semiconductor process engineers and technicians. Superconducting qubit fabrication is cleanroom lithography, and the skills transfer with context.
Control systems engineers. Into calibration and feedback control, where the problem structure is familiar.
Software engineers. Into the quantum software stack, most of which is conventional software engineering.
Physics graduates without quantum specialisation. A short path.
Laboratory technicians from research environments. Underrated. Precision measurement, vacuum, cryogenics and instrumentation skills are exactly what these machines need day to day.
Cryogenic and laboratory hazards. Quantum hardware work involves cryogenic liquids that cause severe cold burns and can displace oxygen in enclosed spaces, high vacuum systems, strong magnetic fields, and laser systems in ion and atom platforms. These hazards require site specific risk assessment, controlled procedures and supervised practical training. Astra Trainer builds technical understanding of the field. It does not provide hands-on hazard training, laboratory authorisation or site qualification.
What to take from this
Quantum computing is staffed mostly by engineers, and the entry requirement most organisations assume excludes the people the field actually needs.
Physical qubit counts do not indicate capability. Error correction consumes many physical qubits per logical one, and qubit quality drives that ratio.
The advantage is specific rather than general, and strongest for simulating quantum systems.
Nobody can responsibly date the arrival of large scale useful machines, and confident predictions should be read as commercial claims.
And the engineering skills this field needs, cryogenics, microwave, precision control, are valuable whether or not the timeline holds, which makes them a sound investment either way.
Do you need a physics doctorate to work in quantum computing?
For core theory, usually yes. For most of the headcount, no. Cryogenics, microwave electronics, control systems, fabrication, software and integration make up the majority of roles and are recognisable engineering disciplines.
Why is qubit count a poor measure of progress?
Because physical qubits are noisy and lose their state. Error correction encodes one reliable logical qubit across many physical ones, and the ratio depends on hardware error rates, so more qubits of worse quality can be further from useful.
Will quantum computers replace ordinary computers?
No. They offer advantage on specific problem classes, most credibly the simulation of quantum systems such as molecules and materials, and on certain algebraic problems. General computation is unaffected.
When will quantum computing be commercially useful?
Nobody can answer that responsibly. The remaining problems are hard engineering problems with real uncertainty, and forecasts in this field have been wrong in both directions.
What should an organisation do now?
Build enough literacy to evaluate claims, check whether its problems fall in the plausible set, develop engineering skills that are valuable regardless, and treat cryptographic migration as the one area where delay has a present cost.