Every big company that has started a quantum computing team runs into the same wall. The machines are getting better every year. The people who know how to use them are not getting any easier to find.

QodeX Quantum has decided not to fight over those people. It makes the ones a company already has count for more.

Quantum computing, in plain words

An ordinary computer works with bits, each one a 0 or a 1. A quantum computer works with qubits, which follow the strange rules of physics at the scale of atoms. For a small set of problems, such as how a drug molecule behaves, how air or water flows around an object, or how risk moves through a financial market, that should one day let a quantum computer do what no ordinary computer can.

The machine on its own is not enough, though. It needs an algorithm, which is simply a recipe telling it what to do, step by step. Good quantum recipes are rare and very hard to find. Today's machines are also noisy and make mistakes, so a recipe that looks perfect on paper often falls apart on real hardware. QodeX says that two out of three published methods fail when they are run on a real processor.

Finding recipes that work is the job of a quantum algorithm team. That is who QodeX sells to.

What they built

QodeX Quantum, based in Chicago with part of the team in Maryland, makes AlphaPsi, an AI research partner for company quantum teams. It reads the research as it comes out, which no person can keep up with any more. It suggests new designs by combining methods that are already proven. It explains the reasoning behind every suggestion, so the scientists can check the argument and not just the answer. And it runs the designs on real quantum hardware, IBM's for now, and records what happened.

Ethan Evans, one of the co-founders, explains why that matters. The large AI models everyone uses were not built for quantum algorithms. Without the right tools they take shortcuts, or give answers that look right and are not what the researcher asked for. QodeX gives the AI proper tools for the job and leaves the human expert in charge. In his words, the idea is not to take the expert out of the driver's seat, but to give them a Ferrari. The aim is not only speed. It is better results.

The founders did not set out to build this. Three of them, Ethan Evans among them, led quantum research for the US Navy. Ethan has a PhD from Georgia Tech on controlling quantum systems, work that grew out of a robotics lab. They left to start a company, joined by Stu Pino, a founder who had built and sold a technology company before. To compete with much bigger teams, they built tools that let a small group work fast and well. One day they realised the tools were the product.

The team is seven people, five with PhDs. They cover mathematics, optimisation, machine learning, and quantum algorithms, with one of them holding an Oxford PhD. Their first employee spent six years at IBM as global programme director for its quantum innovation centres. The company says it has six patents filed and papers in peer-reviewed journals including Physical Review A.

The backing and the early proof

QodeX raised $1.5 million before its seed round, and IBM's venture arm was the largest investor. Ethan says it is one of very few early-stage quantum companies IBM Ventures has backed. The company was one of only seven to graduate from the quantum stream of Creative Destruction Lab in Toronto, out of a class of more than thirty, a programme that removes companies as it goes. It also finished Alchemist Accelerator in San Francisco. Both programmes finished in May and June 2026.

It is running a pilot with one of the best-known names in biotech, using AlphaPsi to speed up research for drug discovery, and it says the work is already beating results that customer had been building on for years. With NVIDIA it has a paper on the way about simulating how fluids flow using quantum computers. On IBM hardware, it says, it has already run a 52-qubit forecast with a circuit that no supercomputer can simulate today. Research agreements with US government laboratories are in progress, and it is in talks with companies in oil and gas and in engineering design software.

QodeX is raising a seed round now. The money is mainly for people: eight to ten hires over the next one to two years, with enough runway for about two years.

Quantum software is a busy corner. Classiq turns a description of what you want into a quantum circuit. Q-CTRL makes noisy machines behave better, and Haiqu squeezes more out of today's hardware. Phasecraft and QC Ware design algorithms for clients, Riverlane builds error correction for the machines themselves, and Quantinuum builds both machines and software. Most of them either sell a tool for one step or do the research on the client's behalf. QodeX sits beside the client's own scientists for the whole cycle, and keeps the result of every run.

The point: they sell the hires nobody can make

According to McKinsey, only one in three open quantum jobs has a qualified candidate. Around 40 per cent of the Fortune 500 already have quantum teams, Ethan says, and demand keeps growing. Those companies cannot hire their way out. So QodeX does not sell software in the usual sense. It sells what the missing hires would have produced: a small team that works like a much bigger one.

Time makes it urgent. A quantum team gets two or three real research cycles a year. Industry roadmaps expect error-corrected quantum machines around 2028 or 2029, which leaves each team six to eight tries to find the method that wins for its business. QodeX says two thirds of that time is lost to reading papers, failed runs and results too weak to learn from. Better hardware will run a good method faster, but it will not tell you which method is good.

The part that grows over time is the record of failures. Every run on real hardware shows which methods hold up and which break, on which machine and why. Most of that never gets published, so no AI model has learned it. QodeX keeps it, and each new design starts smarter than the last. The company is clear that it learns which building blocks work, never the customer's problem or results.

The formula is worth keeping: when a field cannot hire experts fast enough, sell a tool that multiplies the experts it has, and keep the knowledge of what failed, because nobody else writes it down.

Where else this works

This part is opinion rather than anything the company has said.

The same shape fits any field where experts are scarce and failed experiments disappear into lab notebooks. Fusion energy. New battery chemistry. Chip design. Rare-disease drug research. Nuclear engineering. Climate modelling. Materials for aircraft and rockets. In each one, the tool that remembers every failed attempt becomes more valuable with every attempt, and the team that uses it moves faster than one that only reads what was published.

Why it is bigger than it sounds

The short version is an AI helper for quantum scientists. The bigger idea is simple.

  • It fills jobs that cannot be filled. Companies cannot find enough quantum experts, so QodeX makes the ones they have do the work of more.
  • It keeps what nobody writes down. Most failed attempts are never published. QodeX records every one, so its advice gets better with every run.
  • The clock is ticking for its customers. Teams have only a handful of tries before quantum computers become truly useful. Anything that makes each try count is worth paying for.

What to watch

The measure is the biotech pilot. A household name saying, in public, that AlphaPsi shortened its research would open doors across drug discovery.

The second is the NVIDIA paper on fluid flow. Engineering design and energy companies spend heavily on simulating flow, and a result they can check is a reason for them to call.

The third is the seed round and the hires it pays for. Eight to ten more people would more than double the team, and in a field this short of talent, the people QodeX can attract are part of the product.