Poolside Collaborates with AWS to Launch Customized AI Coding Assistant

Poolside has established a strategic partnership with Amazon Web Services (AWS), a renowned public cloud service provider. Through this collaboration, users can access Poolside's foundational models and generative AI assistants via Amazon's Bedrock platform.

Details announced at the AWS re:Invent 2024 conference reveal that customers will be able to utilize these foundational models securely and privately, integrating their own data to create customized coding assistance tools. Earlier, in October, Poolside completed a $500 million Series B funding round led by Bain Capital Ventures, with participation from investors such as Nvidia Corporation and eBay Inc. Poolside focuses on developing a range of generative AI coding models, positioning itself as a competitor to GitHub Copilot.

Poolside's models have been deployed across numerous Fortune 2000 global enterprises, although specific use cases remain limited in public disclosures. These models are trained using reinforcement learning techniques, continuously enhancing performance based on feedback from code execution. Additionally, Poolside offers a comprehensive suite of software toolkits for AI training tasks and tools for evaluating model accuracy.

Amazon Bedrock, as a fully managed service platform, enables developers to access high-performance foundational models from various leading AI companies through a unified API. Leveraging Bedrock, Poolside's models are better positioned to meet enterprise requirements for privacy, security, and compliance, and allow for the deployment of models and training data within the same secure environment.

As part of the collaboration, Poolside plans to leverage AWS's Trainium chips to provide inference support for its Malibu and Point models, optimizing cost efficiency. The Trainium chips are specifically designed AI accelerators that effectively reduce operational expenses.

The two companies have previously collaborated, with Poolside's "learning from code execution feedback" technology already operational on AWS infrastructure, capable of handling up to 10,000 code executions per minute. This technology enables Poolside to continuously enhance model quality by generating large-scale synthetic training data from real-world programming tasks within actual code repositories, all without relying on customer data.

The partnership between Poolside and AWS also aims to streamline the procurement, contract negotiation, and budgeting processes for customers, facilitating the adoption of these models.