There is no surprise for them thinking on these lines, for Saurabh Kumar Vij, is a scientist with experience working for CERN and his brother Gaurav Vij is an artificial intelligence (AI) researcher. Together, they have already developed a business model for making supercomputers available for scientists, engineers, and designers to further their research.
Q Blocks aims to provide access to affordable supercomputer experience and has built on technologies such as distributed and peer-to-peer computing. Such an access will especially be useful for high performance computing applications such as ML model training, running simulations, or big data analytics, among other computer intensive applications.
This secure supercomputing environment is powered by crypto-mining setups and gaming machines.
The founders aim at bringing supercomputing access to more than 5 million data scientists and designers and they expect that this affordability would further expand this market.
Saurabh shares the story of Q Blocks, saying they founded this company “to help the world transition to affordable supercomputing.” Building on the history of how it all came together, he adds that during the gold rush of crypto-mining in the year 2017, many miners invested heavily in infrastructure to mine crypto like Bitcoin and Ether. But, the crash of the crypto space rendered these powerful machines useless and almost 2mn powerful GPUs idle.
“On our way to developing exa scale-level capabilities, our first stop was to make high performance GPUs virtually stackable and remotely accessible by others. Though the idea of using someone’s else’s computer to fund own research is not new in the scientific world, with The Folding@Home project leading the race with our 1Exaflops of computing power,” shares Saurabh.
So why isn’t everyone doing this already? Q Blocks shares that creating an access to a network of distributed GPUs is not an easy task. First, it is required to build a reliable connect to such remote machines and then to provide a secure developer environment and then a capability to manage the GPUs on remote machines.
Q Blocks has gone a step ahead and developed outage protocols. They emphasize on this service being affordable, as the benchmark tests they have done with 60,000 images shows their system performed the same task in 1/3rd the time and at 1/10th the cost of AWS-powered GPU.
Their vision is to expand their network to include Xbox and gaming stations in the future, while their long-term goal is to provide a reliable service to handle problems such as weather forecasting, synthetic data creation for hedge funds, fraud detection, deep fake detection, computational fluid dynamics for governments, among various other things.
The company is on the way to launching subscription packages, using which users can instantly launch their Jupyter Notebooks with pre-configured AI frameworks. This would include information on learning and testing more data science models.
Going forward, the focus markets for Q Blocks are US, Canada and India – especially Bengaluru in India, where they have connected with multiple data scientists and ML engineers.
Q Blocks is one of the ten startups that successfully completed the mentorship-driven Techstars Bangalore Accelerator's 2020 programme recently. “Techstars Program helped us in meeting people and companies which would have otherwise taken years to connect with. Helped us for an early beta. They also gave us mentors we are still working with. Some of the mentors helped us in bringing more clarity about the B2B environment and norms we should follow,” added Saurabh. “It's truly an accelerator from a networking and growth standpoint.”
The intensive programme provides support to the startups through an investment of $120,000 in each company, mentorship, and support resources as they develop and refine aspects of their businesses to enhance product-market fit and position themselves to scale.