Zapata: Solving intractable problems
If you had invented the Natural Language engine behind Apple’s Siri you might be inclined to rest on your laurels. Not so for machine learning expert Christopher Savoie, CEO and Co-Founder of Zapata Computing. After running data and analytics for Nissan, Christopher launched a photonics company using machine learning capabilities emanating from Alan Aspuru-Guzik’s group at Harvard. Based on that experience, in 2017 Chris and Alan decided to spin out Alan’s quantum information group into what is now Zapata.
“At a big picture level, we are focused on problems that large organizations face that are difficult, if not impossible, to solve via classical computing. Take chemistry as an example. Chemicals function as quantum entities, as wave functions, which are impossible to exactly quantify using classical computing. Classical computing can only approximate their description and properties. And the more complex the chemical, the less accurate the estimation. Quantum computing, on the other hand, can precisely model those properties, therefore providing game changing advantage.”
A second example which Savoie gives is what’s often referred to as ‘the traveling salesman problem’, one of optimization within route scheduling and supply chain management. The parallel computation advantages of quantum computers enable an exponential speed-up to tackle these. “It’s like trying to escape a maze. With classical computing, you can only proceed one step at a time. With quantum computing you can zoom out and see the whole maze, and plot an immediate path out. Just taking supply chain management, imagine shaving off just three-quarters of one percent of supply chain costs for a large organization. That is a huge saving, not just in dollars, but in the amount of carbon we are putting in the atmosphere. Imagine if a large logistics provider like FedEx or UPS could truly optimize it’s aeroplane & delivery network? That would be a ground breaking.”
Several quantum computing start ups have also identified these opportunities, and are trying to tackle these problems. What sets Zapata apart from the rest of the field?
“It stems from the uniqueness of our team. Alan is a true pioneer in this field. He came up with the first practical application of quantum computing to the real world of material science. Because of that innovation, back in 2004, he attracted students who wanted to work on real world problems at the frontier of chemistry and material science. As a result he ended up training many of the theorists who are working on practical applications at the leading companies in the world.”
Fast forward to 2014, when experimental groups around the world started using actual working qubits.
“What John Preskill has termed NISQ-era machines, not perfect sure, but still working. Alan’s group came up with the very first algorithm, a variational algorithm, that could run on these noisy qubits, called VQE. Every single algorithm that runs on a NISQ machine is a variational algo, and it is a derivative of that original methodology. Four of our scientists are actually the top post-docs that actually wrote those original algorithms. Because of this we are able to attract the top variational algorithm experts in the world. People who can write practical algorithms that can run on the 50 qubit, 100 qubit machines we have today.”
The number of people who can write original, novel variational or heuristic algorithms is very limited worldwide. And Zapata is fortunate enough to have the best of those people working within it’s walls. People who can say that they are first authors on novel variational algos, a rare achievement.
That being said, Savoie identifies that hiring and retaining that level of talent is still a key challenge as a start up. Another challenge, one that is out of his control, is the hardware development.
“But there is so much investment going into the hardware, the smart money wouldn’t bet on the nay sayers! It’s just a matter of time,” says Savoie.
“Our mission is to create the software platform and tools to best take advantage of that hardware when it does supersede classical computing. Already we are creating useful platforms for 100, 200, 300 qubits.”
Where amongst the enterprise base is he seeing the early successes coming from?
“Chemistry is such a no brainer for quantum computing. Applying quantum computing is really the only way to properly understand chemical properties. So as a result, some of our early engagements are in the chemistry area because those companies, they get it. They can see what impact this will have on their business and want to be ready for it.”
“Next in line are companies with the logistics challenges we talked about. That’s a solution that can apply across different industries, and government, so it needs a degree of customization and domain expertise across different verticals. But again, the impact in terms of value is tremendous. Currently, with classical methods, we just can’t optimize to the extent that quantum computing will allow us to.”
However, there is still plenty of work that remains to be done.
“We still need to continue adapting the algorithms to be more and more robust to noise, and make them do more with fewer qubits. For example, CUSP, a new algorithm that we released with Google, allows for the quantum computer to teach itself with machine learning mechanisms to run on fewer qubits. It makes it much more efficient. For the foreseeable future we will be trying to run even more computation on fewer qubits,” explains Savoie.
“We are so busy with customer engagements, and actually designing for production. With the experience we are gaining from customer engagements, we carry on building out our platform too. The goal is to then make the platform open and available for other developers to utilize, so they don’t have to reinvent the wheel. The customers we are working with are global fortune 100 companies. These early adopters are committing time, money and resource to make these applications happen, and realize the value from them. They can see the potential value that quantum computing can deliver. That’s what’s truly exciting for us!”
For questions or feedback, please contact Amit Das: firstname.lastname@example.org
To visit the Zapata website: https://zapatacomputing.com/