Riza has emerged from stealth mode with a bold vision: letting LLMs not only write but also run code instantly in production. The startup raised $2.7 million in seed funding led by Matrix Partners, alongside contributions from 43, to support this mission.
The goal? To help developers and AI agents generate, execute, and test code in real time, safely and reliably.
Riza Solving Pain Points in AI-Driven Development
Founders Andrew Benton and Kyle Gray — former engineers at Stripe, Twilio, and Retool — started Riza to fix a big problem. While LLMs can write code well, executing it safely has been a major hurdle for developers. Running unreviewed, dynamically generated code brings huge risks like security vulnerabilities, data leaks, or infrastructure damage.
Typically, teams must set up complex infrastructure or rely on human reviews to run LLM-generated code securely. This slows development and limits how AI can be used in software projects.
Riza’s breakthrough came from a real-world challenge. A former colleague needed a way to run AI-generated code safely, without constant manual review. Benton and Gray built a prototype using WebAssembly (WASM) from their open-source tool sqlc — and it worked. That prototype became the foundation for Riza’s product.
Now, Riza provides “AI-first infrastructure” through a secure WASM-based runtime. It supports multiple languages, including Python and JavaScript, and isolates code from the host environment. This makes it easy for developers to execute untrusted code without worrying about complex setup or exposure to security risks.
With this system, developers can run code in development, CI pipelines, or even production environments — safely and efficiently. Riza’s platform also reduces latency and increases the flexibility of AI agents by allowing them to create and run their own tools.
Introducing Just-in-Time Programming
Riza coined the term “Just-in-Time Programming” to describe its innovative approach. This model lets AI systems create and execute code as needed. It offers massive potential, but also comes with real security risks like unauthorized access or system compromise.
Companies using Just-in-Time Programming either need to build strong protections or rely on Riza’s secure platform. One example: a customer uses LLMs to generate code that pulls and analyzes data from multiple systems. That code then builds reports, complete with charts and insights, based on live data. The AI couldn’t perform these tasks directly — but it could write code that could.
Now generally available, Riza’s platform is seeing major growth. In March alone, users submitted more than 850 million code execution requests using Riza.
Patrick Malatack, a partner at Matrix, calls this a “generational shift” in how developers and AI will use computing power. “This doesn’t replace humans,” he said, “but it creates brand-new needs that old infrastructure can’t handle.”
With their deep experience building developer tools, Riza’s team is poised to shape the future of AI-powered software development.
