Kuano

Kuano is a next-generation biotech company using AI and quantum mapping to solve big problems in the drug design space. Its mission is to design high-quality drug candidates rapidly and cost-effectively for targets that are hard to drug or even deemed undruggable.

Kuano’s approach combines state-of-the-art physical simulations with AI to go beyond the limitations of both methods when used alone. As the Kuano platform requires input from its own scientists, it isn’t available as a service to be used independently. Instead, Kuano is building collaborative partnerships where its technology can be embedded into a partner’s drug design projects, helping to solve ongoing problems of potency, selectivity, or drug resistance.

To power its platform, Kuano was using a combination of high-performance computing (HPC) and local hardware solutions, as well as graphics processing units (GPUs) and storage solutions from multiple cloud providers. However, these solutions were proving to be expensive and there were issues with the customer support and documentation provided.

With a growing customer-base and an increasingly variable workload, Kuano needed to find a new provider that could offer flexible, reliable solutions at a competitive price. This would enable Kuano to scale and deliver a quality service, whilst also controlling its cloud consumption costs.

After comparing a range of cloud providers, Kuano chose OVHcloud based on the latter’s diverse portfolio of Public Cloud solutions, which offer the performance required to handle AI workloads, on an affordable price plan.

To train its AI models and run scientific simulations using HPC, Kuano adopted NVIDIA Tesla V100 and V100s GPU instances, hosted on OVHcloud’s Public Cloud. Designed to support AI, machine learning and HPC processes, the Tesla V100 and V100s deliver the power of 100 central processing units (CPUs) in a single GPU. As the first GPUs to exceed the 100 teraFLOPS (TFLOPS) barrier of AI training performance, the Tesla V100 and V100s enable data scientists to train AI models faster - in a matter of days, as opposed to weeks.

As an AI-based solution processing sensitive scientific data, the Kuano platform also required scalable, secure data storage. To meet this requirement, Kuano adopted OVHcloud Block Storage and Object Storage, both hosted on the Public Cloud. 

Block Storage is a highly scalable and accessible storage solution. It stores data in equal-sized blocks with a unique identifier for efficient data retrieval. Capacity can be increased by adding extra storage disks, making it ideal for storing the large amounts of data required for AI platforms, such as Kuano’s. As the volumes are triple-replicated on three separate disks and hosted in multiple datacentres, this ensures data durability, and users can also take snapshots of their data at any time to retrieve later.

With unlimited storage capacity, Object Storage is also well-suited for storing AI data. It organises data into objects containing a unique identifier and metadata, allowing for easy access, retrieval and management. Object Storage also has a range of data security and compliance features, such as Object Lock, Access Control List (ACL) Management, and Server-Side Encryption (SSE). These features add an extra layer of confidence for AI companies like Kuano, that need to ensure data is processed in accordance with data laws and regulations.

As all the GPU and storage solutions were hosted on OVHcloud’s Public Cloud, Kuano was able to take advantage of transparent pay-as-you-go monthly billing and a reliable, resilient and highly-available infrastructure. In addition, as OVHcloud’s cloud services are ISO/IEC 27001, 27017, 27018 and 27701 certified, Kuano could rest assured that its data was safely secured and compliant with healthcare and pharmaceutical industry standards.
 

“The wide variety of instance types combined with the monthly subscription option has allowed us to work very efficiently.”

Alex Punter, Computational Drug Design and ML Scientist at Kuano

Adopting ultra-high-performance NVIDIA Tesla GPUs enabled Kuano to build and deliver its platform much more efficiently. As these GPUs are hosted on OVHcloud’s Public Cloud, Kuano was able to benefit from a highly available and reliable cloud, with a simple interface and multi-local datacentre infrastructure. This enabled Kuano to access its GPUs 100% of the time, whereas the company’s previous cloud provider was only able to offer 70-80% availability.

After moving to OVHcloud, Kuano also made a major 60% saving on its running costs. This was due to the Public Cloud’s predictable and transparent monthly pay-as-you-go billing, which enabled Kuano to control its cloud costs effectively. Also, despite Kuano’s flexible workload, the company was able to benefit from committed-use discounts across 90% of the services it adopted from OVHcloud. Kuano’s previous providers only offered such discounts over a period of years – an approach incompatible with the varied nature of Kuano’s work.

OVHcloud’s Block Storage and Object Storage solutions delivered further benefits, providing Kuano with scalable, durable and accessible storage for its large AI datasets. The extra security features also ensured Kuano was able to store its data in compliance with healthcare industry standards. In addition, OVHcloud’s responsive customer support and up-to-date documentation enabled Kuano to remedy issues quickly and continue its mission to design drug candidates using the power of AI and quantum computing.

“We found OVHcloud’s documentation to be concise and up-to-date. We also found OVHcloud’s support to be responsive and helpful, with OVHcloud’s costs being very reasonable. In particular, the monthly committed-use discounts allowed us to adapt our costs according to business needs, whilst still obtaining good value for money. This has allowed us to make substantial savings compared to our experience with other providers. This combination of economy and ease-of-use has enabled us to make a huge amount of progress in our work, ranging from simulations of quantum computers to models of enzyme reactions, as well as high-throughput generative AI.”

Alex Punter, Computational Drug Design and ML Scientist at Kuano