QClairvoyance's quantum chemistry bet: Transitioning from research to pharma partnerships

Quantum computing is starting to impact molecular simulation, potentially helping pharmaceutical companies improve early-stage drug discovery. Published this month in the Journal of Chemical Physics, a study by QClairvoyance Quantum Labs revealed that breaking eight small molecules into smaller computational tasks allows a quantum computer to reproduce chemically accurate results. However, the findings are an initial proof of concept and not yet a breakthrough in drug discovery. In this conversation with The Economic Times Digital, Col. Sai Shankar P, VSM (Retd.), founder and director of the Secunderabad-based QClairvoyance Quantum Labs and former technical manager to the National Quantum Mission, discusses the study's significance, the road to pharmaceutical applications, and how the deep-tech startup plans to turn its research into a business. Edited excerpts.

ET: What problem does this research address, and what did the team demonstrate? Why were these eight molecules selected, and how were the results checked?

Sai Shankar P (SSP): In short, our study shows that today's quantum computers can already help with real chemistry, as long as the problem is divided into pieces small enough for them to handle.


The challenge is this. Every medicine works because of how its molecules behave, and that behaviour follows the rules of quantum physics. Ordinary computers can only approximate it, and the approximation weakens as molecules get bigger. Quantum computers should be the right tool, but today's machines are still small and prone to errors, so they cannot handle a whole molecule at once.

So we broke the job up. We cut each molecule into small pieces, solved every piece on a real IBM quantum computer, and joined the answers back together. For all eight molecules, the results met the chemical accuracy threshold, which is the precision standard chemists use before relying on a prediction. It is the difference between a map that gets you to the right street and one that only gets you to the right city. Most of our results were well inside the minimum requirement.

We chose eight small molecules that are simple enough for today's chips but not artificially easy. Earlier work used neat, symmetrical shapes. Ours are lopsided, which is closer to what drug makers actually work with. A couple are familiar in medicine as well. Urea type structures appear in a number of medicines, and methoxyamine has been tested in early cancer trials.

To check the answers, we compared each one with an exact calculation done on an ordinary computer for the same pieces. That shows the quantum step is working as it should. It is not yet a comparison with lab measurements, and that is the standard we are working towards.


ET: What is QClairvoyance's specific contribution to this study compared with earlier work using these methods?

(SSP): We took a method that others pioneered and tested whether it holds up on harder, more realistic molecules. The original idea belongs to a team from the Cleveland Clinic, Michigan State University and IBM, who first showed that a molecule can be split and a quantum computer used on the pieces. They worked with neat structures such as a ring of hydrogen atoms.

We added three things, the first being, lopsided molecules. With no symmetry to lean on, it is harder to decide which parts matter. This is much closer to real drug chemistry.

We also took a closer look at one key setting. When a molecule is cut, we decide how much of each piece's neighbourhood to keep. For most pieces it made little difference. For one, the sulfur atom in HOSCN, it mattered a lot, and our most detailed setting gave a worse answer than a middle one. It works like packing a suitcase: take too little and something is missing, take everything and it becomes too heavy to carry. More detail is not always better.

Last, we used a shortcut. We run a quick check on an ordinary computer first to see where that setting will matter, before spending expensive quantum time.

We have shown this on one molecule so far, and the next step is to test it across more. It is important to be precise about which part is ours, and there is more work to do to understand how broadly the findings apply.


ET: Why were quantum and conventional computers used together? How did dividing the calculations into smaller tasks help address current hardware limitations?

SSP: Each machine is good at something different, and today neither is enough alone.

A quantum chip can explore many possibilities at once, but it makes mistakes. An ordinary computer is dependable, but it struggles as molecules grow. So, the quantum chip acts like a scout. It points out which electron arrangements are most likely to matter, and the ordinary computer cleans up the report and does the heavy calculation. Most of the heavy lifting still happens on the conventional side, and that is an accurate picture of where the field stands today.

Cutting the molecule into pieces matters because mistakes pile up as a quantum programme gets longer. Solving HOSCN as a whole would have needed 50 qubits, the basic building blocks of a quantum computer. By treating each atom as its own piece, the biggest piece needed 30 and most needed 20 or fewer. Modern ships are built the same way, in blocks joined at the end, because each block is easier to build and check than the whole hull.

Experience in large operations points the same way. One unit is never asked to do everything and the job is broken down and each part goes to whoever is best placed to do it. There is one catch here. Joining the pieces can add small errors of its own, and our checks do not measure them. That needs a closer look.


ET: What do the reported accuracy and reduced qubit requirements establish? Did the study demonstrate any advantage in total computing time or cost over conventional approaches?

SSP: They show that the method works on real hardware today. A real quantum chip, with all its noise, can take part in a chemistry calculation and reproduce a trusted answer. Splitting the molecule also made the quantum job smaller.

They do not show an advantage in time or cost, and we are not claiming one. At this size, an ordinary computer can solve each piece exactly in moments, so there is nothing to beat yet. The quantum chip itself ran for roughly one to three minutes per molecule, but the full set took about two days, mostly waiting in the queue for shared hardware, and the work on ordinary computers comes on top of that. We have not compared costs.

Think of it as a successful test flight. It shows the plane can fly. It does not show that it is faster than the train. That question gets answered when molecules become too large for ordinary computers, which is where quantum is expected to matter. For now, the priority is to establish the method's reliability and test it on larger, more complex molecules.


ET: Where could this approach eventually fit into pharmaceutical R&D? How far are these molecular calculations from identifying or testing a potential medicine?

SSP: At the very start of the process, and a long way from a medicine.

Before anyone makes a new molecule in a lab, chemists want to predict how it will behave and how well it will attach to its target in the body. Developing a new medicine commonly takes 10 to 15 years and billions of dollars. Many candidates fail late, partly because early computer predictions are only approximate. Better early predictions would let weaker candidates be dropped sooner and stronger ones move ahead with more confidence.

Big drug makers are exploring this area too. Boehringer Ingelheim announced a multi-year partnership with Google's quantum team in 2021 to explore exactly this. Given India's established pharmaceutical industry, there is a real opportunity for Indian researchers and companies to contribute to this emerging area.

There is still considerable work to be done before this approach can contribute to drug discovery. Our eight molecules are tiny. The heaviest weighs about 75 daltons, and aspirin, a very simple medicine, is more than twice that. None of the eight is a drug, and our calculations do not yet include a protein or a measure of how strongly a molecule binds to its target. In a separate study on a chip from the Finnish company IQM, we looked at a small slice of amantadine, an approved medicine used against flu and Parkinson's disease. That was also a computational test, not a step towards testing a drug. What we have shown is that one important building block works.


ET: Beyond publication and peer review, has the work been independently reproduced or evaluated? What further validation is planned?

SSP: The work was published in the Journal of Chemical Physics on 19 October 2026, after being submitted in April and accepted in September. Independent experts examined it before it went out. That is not the same as another group repeating the experiments, and as far as we are aware, no outside group has done so yet. We would welcome it, because a result earns real trust when others can repeat it.

It is also right to be open that two of the authors work with us as paid consultants, and the paper says so clearly. In one of our earlier studies, the results stopped converging the way we expected. The team stopped and asked why it was happening, went back, rebuilt part of the workflow and continued until the results were consistent. That experience taught us to examine unexpected results carefully.

On further validation, our next steps are to repeat the threshold study on all eight molecules, since so far we have done it only for HOSCN, and to test larger basis sets and molecules with unpaired electrons, because our current comparison uses the simplest basis set so that we can check against an exact answer.

We also need to compare against laboratory measurements, since today we compare only against another calculation, and to test against the best conventional methods on a question chemists care about, such as how strongly a candidate binds its target. The data behind the study are available in the article and its supplement, and we will support any group that wishes to repeat the work.


ET: Has QClairvoyance begun engaging pharmaceutical companies or research institutions to explore applications? Are these engagements at the discussion, collaboration, pilot or paid project stage?

SSP: With research institutions, yes, at the collaboration stage. The authors of the paper include researchers from IIT Bombay and the University of Arizona. We have research partnerships with CSIR's IICT, TIFR, IIT Bombay and Ahmedabad University. In September we also signed an MoU with Samgnya, the national hub set up at IIT Madras under the National Quantum Mission, covering research, validation and eventual deployment.

With pharmaceutical companies, we are at the beginning. We have shown the technology on small molecules, and the next step is to find pharma and research partners who will test it on problems they care about. We do not have a pharma partnership or paid project to announce today, and we would rather say so than overstate it.


ET: Has this work resulted in proprietary software, intellectual property or patent applications? How does the company distinguish its own developments from the existing methods used in the study?

SSP: A clear line between what is ours and what is not is important.

The core methods in this study, splitting a molecule into pieces and using a quantum computer to sample electron arrangements, come from published work by others, and we built on open scientific software. The paper does not claim a new algorithm. What is ours is how we applied and tested these methods on lopsided, low-symmetry molecules, the study of how much of each piece’s surroundings to keep, and the idea of a quick pre-check on an ordinary computer to see where those setting matters. All of this is published openly in the paper.

Beyond this study, our team is working on related methods, including machine-learned ways of selecting the most relevant electron arrangements, which are described in separate preprints. We have also filed a patent on other aspects of our work, which are not part of this paper. Because the application is pending, we are not able to discuss its details at this stage.


ET: How does QClairvoyance plan to build a business around molecular simulation: through research services, software licensing, joint development or another model? What evidence would a customer need before adopting the offering?

SSP: For now, we plan to work alongside pharma companies and research institutions on specific problems, using the molecules they care about, and learn together where quantum adds real value. Whether that becomes services, software licences or joint development will depend on what the results show, and that decision should rest on evidence.

At the moment, this is a research partnership and not a product, and we will say so. For perspective, QSimulate in Boston, which uses quantum mechanics-based modelling on ordinary computers, reports working with five of the world's top 20 drug makers. That tells us the industry will work with firms that offer better molecular modelling once it is proven useful.

ET: Is the molecular simulation business currently generating revenue, or does it remain at the research stage? How is the company supporting its development during this period?

SSp: The company is at the research stage and has raised Rs 30 crore from MSN Nextech Ventures, with a team of about 40 people. For this study, IBM gave us hardware time through its Quantum Startups Program. Our other lines, quantum safe cybersecurity and training, help carry the company while this work matures.

Deep technology research takes time to mature, and we are fortunate to have investor support as we continue developing this area of work.


ET: What measurable milestones is QClairvoyance targeting over the next 12 to 24 months, and how does molecular simulation fit within its broader business across cybersecurity and education?

SSP: Our direction is clear, and specific targets will follow the results. We want to move to larger molecules and more detailed models, and turn our research into collaborations with pharmaceutical companies.

In the cybersecurity area, we want to deepen our work with banks, insurers, payment networks and exchanges. The government's quantum safe task force has recommended that critical sectors begin formal implementation of post quantum security by 2027. That falls within most organisations' current planning cycle, so the time to begin is now.

In the sphere of education, we want to train more students and professionals. We have trained over 1,000 so far.

For organisations beginning to prepare, a good first step is to list where encryption is used across their systems. You cannot protect what you cannot see.

The simplest way to describe our business is one near-term business and one long-term bet. Cybersecurity has a clear planning deadline, so many organisations are beginning to prepare. Molecular simulation is the longer road, and that is why we are being patient with it. It is better to take a little longer and make something work properly than to announce it before it is ready.