Where we help
- Readiness assessment. Which of your computational problems, if any, are candidates — and what would have to be true for them to become so.
- Algorithm and hardware suitability. Which approach fits the structure of your problem, and which platform is plausible for it.
- Resource estimation. Qubit counts, circuit depth and error rates required, set against what hardware roadmaps actually promise.
- Proofs of concept. Working circuits on real machines, benchmarked against a serious classical baseline.
- Post-quantum cryptography. Inventory, risk assessment and migration planning for the one deadline that is already running.
- Training. Bringing your engineers up to a level where they can read the literature and judge vendor claims themselves.
What is credible today
Quantum computing is real, useful in a narrow set of places, and years from being useful in most of them. A short version of where the evidence sits:
- Chemistry and materials simulation is the strongest long-run case, and the one that most clearly needs error-corrected hardware. Current roadmaps put that in the 2030s.
- Optimization is the most oversold. No quantum method has convincingly beaten a well-tuned classical solver on a real instance.
- Machine learning runs the other way: AI is already load-bearing inside quantum computers — error decoding, calibration, compilation — while quantum has yet to improve a production model.
- Cryptography is the exception with a live deadline, because encrypted traffic captured today can be decrypted later.
The cryptography deadline
This is the piece of quantum work almost every organization should already be doing. Data you send today under RSA or elliptic-curve encryption can be recorded now and decrypted once a sufficiently large quantum computer exists. If your secrets must stay secret for a decade or more, the clock started before the hardware did.
The work is mostly logistics rather than cryptography: inventory what you send and how long it stays sensitive, find out which vendors have a post-quantum roadmap, and migrate the long-lived, externally visible traffic first. The standardized algorithms already exist.
How an engagement runs
An assessment of a few weeks, fixed fee, ending in a straight recommendation: build, wait, or drop it. Roughly half end in "wait" or "drop it", which is a result worth having early. Where it is justified, a proof of concept of one to three months follows, on real hardware and benchmarked honestly. Longer embedded work is available where a team wants the capability in-house.
And AI?
Our other practice is AI consultancy, which is where most organizations should be spending their attention today. We are happy to say that even when it means a smaller quantum engagement.