TORONTO — May 23, 2024 — Agnostiq today announced an update to Covalent, its advanced computing platform. With the release of new capabilities called Function Serve, Covalent becomes an enterprise-grade solution for AI development and productionalization that provides serverless AI model training, tuning and inference in a single platform.
Customers can now take any open-source or proprietary model, fine-tune it, host it, and serve it with complete infrastructure abstraction, all from a single python notebook.
With the explosion of GenAI and LLMs, many startups and enterprises struggle to efficiently train and deploy models due to the lack of tool standardization as well as the complexity of the required computing infrastructure. By providing complete infrastructure abstraction and a unified framework for end-to-end AI development, Covalent provides users a blank canvas to create any AI application they envision, making the productionalization of AI accessible to startups and enterprises, allowing users to transform concepts into ready-to-ship products, and significantly reducing time to market.
“Agnostiq is solving the hardest part of AI application deployment for any enterprise or start-up, which is orchestrating the accelerated compute infrastructure and AI development pipeline,” says Oktay Goktas, CEO of Agnostiq. “Covalent gives developers unprecedented speed and agility to build, iterate, and deploy anything that they can imagine, from multi-agent, multi-modal AI applications, to digital twins for real world simulation. What used to take months with dozens of best-in-class engineers can now be accomplished in hours from a single Python notebook.”
Key features and functionalities of Covalent include:
- End-to-end development: Covalent is a comprehensive AI development platform, allowing users to manage the full lifecycle of productionalizing AI applications, from model training to deployment.
- Infrastructure abstraction and orchestration: Covalent provides complete infrastructure abstraction, eliminating the need for DevOps and AIOps type infrastructure work, allowing developers to focus on code.
- Serverless architecture: Covalent infrastructure is entirely serverless, allowing dynamic resource allocation for the full utilization of expensive and scarce compute resources, resulting in significant cost savings.
- Multi-agent orchestration: Covalent includes native workflow orchestration functionality, enabling users to build production-grade AI agents in a scalable and cost-effective way. (See example here)
Since launching in January of 2022, Covalent has focused on providing users with simplified and scalable access to cloud-agnostic high-performance computing infrastructure, including CPUs, GPUs, and application-specific compute modalities. With Function Serve, Covalent builds on those capabilities and introduces critical new use cases and applications into the platform, particularly for AI model training, fine-tuning and inference workflows. Other use cases of the Covalent platform include large-scale simulations, digital twins, robotics, and any task that requires specialized or high performance computing resources such as GPUs, ASICs and quantum computers.
Learn more about Covalent for AI in this blog. Sign up today and get started with up to 30 hours of GPU compute credits.