(Cyber Magician/Shutterstock)
A new platform from the Palo Alto-based startup aims to level the playing field for AI development. Monster API has launched a platform claiming to be “his Airbnb for GPUs” and has secured $1.1 million in pre-seed funding.
Generative AI is gaining worldwide attention, with applications rapidly being developed for everything from content creation to code generation. A recent Gartner study found that by 2026, 75% of newly developed enterprise applications will incorporate AI or ML-based models, up from less than 5% in 2023. I was. However, the process of developing AI applications using machine learning is expensive and complex. It is expected that many companies will try to join the AI boom.
But what if you could run modern AI models on a cryptomining rig or on an Xbox? I have a question. His brother and co-founder Saurab Viji, an elemental particle physicist at CERN, recognized the potential of distributed computing and contributed to projects such as: [email protected] and [email protected].
Inspired by these experiences, the brothers say they sought to harness the computing power of consumer devices such as PlayStations, gaming PCs, and cryptocurrency mining rigs for training ML models. The company says the brothers have gone through multiple iterations to successfully optimize these consumer GPUs for ML workloads, leading to a 90% reduction in his monthly cloud costs at Gaurav Vij’s startup. I claim.
“We jumped with excitement and felt we could help millions of developers building AI just like us,” Gaurav Vij said in the release.
Monster API allows developers to access the latest AI models such as Stable Diffusion, Whisper AI, LLaMA and StableLM through a one-line command interface with highly available REST APIs without complex setups claim to do so. The platform also has a code-free fine-tuning solution that specifies hyperparameters and datasets.
Monster API’s development stack consists of an optimization layer, a compute orchestrator, a “massive” GPU infrastructure, and a pre-configured inference API. Additionally, according to Monster API, containerized instances come pre-configured with his CUDA, AI frameworks and libraries for a seamless managed experience.
The company claims that its distributed computing approach gives it on-demand access to tens of thousands of powerful GPUs, such as the A100, to train breakthrough models at significantly lower costs. The platform allows developers to access popular underlying AI models at one-tenth the cost of traditional cloud providers such as AWS, the company claims. In one case, Monster API claims that an optimized version of its Whisper AI model running on its platform reduced costs by 90% compared to running on AWS.
“By 2030, AI will touch the lives of 8 billion people. When it comes to Monster APIs, our ultimate wish is to empower developers to unleash their genius and innovate in a matter of hours. Surprise the world by helping,” said Saurabh Vij. “You don’t have to worry about managing GPU infrastructure, containerization, setting up Kubernetes clusters, scalable API deployments, and you also get cost savings benefits.”
“One early customer saved over $300,000 by migrating ML workloads from AWS to Monster API’s distributed GPU infrastructure,” continued Saurabh Vij. “This is the breakthrough product developers have been waiting for. It is a platform that is not only very affordable, but also intuitive to use.”
The Monster API bills based on API calls, not “pay by GPU hours”. The API will automatically scale to meet increased demand, and the distributed GPU network will enable geographic expansion, the company said.
(sauce: twitter)
Access to distributed GPUs could be attractive, as some have suggested that a shortage of GPUs is impacting AI adoption. Quora CEO Adam D’Angelo recently tweeted In his opinion, the AI boom is underestimated due to GPU/TPU shortages.
“This shortage has created all sorts of restrictions on product deployment and model training, but these are invisible. Instead, what we are seeing is Nvidia prices skyrocketing If supply meets demand, things will accelerate,” he wrote. Elon Musk agreed, adding that many other neural network accelerator chips are also in development.
Monster API, on the other hand, says it can provide developers with a scalable, globally accessible, and affordable platform for generative AI by leveraging distributed computing resources. The company offers free trials and subscriptions, which you can view here. If you’d like to learn more or join the developer community, you can join his Discord at the company.
“Generative AI is one of the most powerful innovations of our time, with far-reaching implications,” said Sai Supriya Sharath, Managing Partner, Carya Venture Partners. “It is very important that small businesses, academic researchers, competitive software development teams, etc. have the ability to leverage it for the benefit of society. Monster APIs are needed to thrive in this new world. It provides access to a great ecosystem.”
Related
