James Russo is an engineering lead at HeyGen and co-creator of HyperFrames, an open-source framework that turns agent-written HTML, CSS, and JavaScript into finished videos. He leads development of HeyGen’s programmable video platform, including its API, Model Context Protocol integrations, command-line tools, and HyperFrames.
Russo graduated from Cornell University in 2017 with a computer science degree and a concentration in machine learning. His early career included full-stack engineering at Rockets of Awesome and a first-engineer role at workplace-safety startup Iterate Labs. At Brex, he worked on business accounts, internal developer infrastructure, and bill payments; his infrastructure projects included a Go command-line interface, a Backstage developer portal, and a microservices-testing platform.
Promoted to staff engineer in 2022, Russo became the technical lead of Brex’s backend developer-platform group and shared its approach to scaling engineering with Backstage. He subsequently worked at Empirical Health before joining HeyGen, where his focus shifted from infrastructure for human developers to creative tools usable by coding agents.
The ideas behind HyperFrames
- HTML-native video creation. Russo designed HyperFrames around familiar web technologies and minimal timing metadata instead of specialized video languages. He tested the format against Gemini 3 Flash: if a smaller model could produce workable compositions, stronger coding agents could build on the same foundation. The resulting framework accommodates SVG, Three.js, shaders, WebGL, and other browser-native visual techniques. His AI Engineer World’s Fair talk connects these architectural choices to editable, agent-generated launch videos.
- Deterministic browser-based rendering. Browsers load fonts, images, and animations asynchronously, while video export demands complete, predictable frames. Russo’s rendering architecture freezes the browser clock, advances through individual moments, waits for assets to settle, captures each frame, and encodes the sequence. It also addresses Chrome’s compositor, seekable animations, and embedded-video preprocessing.
- Creative judgment as an engineering problem. HyperFrames supplements code generation with skills for composition, motion, branding, and storytelling, while preserving human control over storyboards, keyframes, and final edits. Russo also maintains video-understand-skills, extending his interest from generating video to helping agents interpret it.
- Engineering judgment over prompt tricks. Russo argues that model selection and feedback loops matter more than elaborate prompts. His writing about coding agents emphasizes context, prioritization, reviewable changes, verification, and deciding what should be built. His Stripe Projects integration applies that approach to autonomous product launches, enabling agents to provision services, handle payment, and generate launch videos through HeyGen’s API.
