Accelerate Government Innovation with Secure, AI-Enabled Models for Software Development

Frontier Model Company Focused on the SDLC
poolside builds frontier AI models purpose-built for software engineering. poolside’s foundation models are trained from scratch using Reinforcement Learning from Code Execution Feedback (RLCEF), a breakthrough approach that generates and learns from synthetic data to deliver superior performance on coding tasks. At the core of poolside is the Model Factory, proprietary infrastructure that transforms model development from slow, artisanal work into a system of rapid, compounding improvement.

Bringing The Model to the Data, Instead of the Data to the Model
Designed for mission-critical environments, poolside models can be deployed fully on-premises—including in air-gapped systems with zero cloud access—ensuring code, IP, and infrastructure remain fully secure. No data ever leaves the mission. Poolside’s quantized models can be installed on workstations that are deployed into the field and highly regulated environments.

Driving Developer Productivity to Support the Mission
Public sector organizations leverage poolside to accelerate every stage of the software development lifecycle. From code generation and test writing to refactoring, documentation, and beyond, poolside enables faster, more secure, and more effective software delivery. Developers can access this intelligence layer through various poolside form factors; IDE extensions (VS Code, Visual Studio, IntelliJ, Coder), Web Assistant, or various APIs.

Data Processing and Model Safety
During training poolside excludes copyleft-licensed code. While other companies may choose to include GPL, AGPL, and other copyleft-licensed repositories in their training data, poolside recognizes that this creates real issues. Poolside is a code-first model provider with training data strategies that differ from general-purpose model providers. These different strategies promote ethical and targeted choices in regards to training data. In addition to data filtering, poolside invests heavily in generating synthetic training data through the proprietary RLCEF (Reinforcement Learning from Code Execution Feedback) technique. This means a significant portion of the training comes from high-quality code that is synthetically created. Not only does this synthetic data help teach the model correct coding patterns and best practices, but it also reduces the reliance on publicly available code altogether.

Flexible Deployment Options
The poolside models are designed to be securely deployed within your organization's security boundary, whether this be in your VPC, utilizing AWS Bedrock, on-prem, or in a disconnected air-gapped environment.