SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Salary: USD 180,000 - 240,000 / annual
HeyGen is building AI-powered video creation technology to make visual storytelling accessible to all. The Video Agent team owns an end-to-end multi-step AI agent that transforms prompts, documents, or assets into finished videos. The agent plans storyboards, writes scripts, selects avatars and voices, calls multiple tools, and integrates with the rendering pipeline. Customers use it daily for training videos, product explainers, marketing content, and more.
In this role, you will improve the Video Agent's ability to execute user intent and validate improvements with data. You will own features from conception through launch and iteration. Your work spans the full stack: designing agent tools, prompts, and context; building evals and running A/B tests; shipping user-facing features in TypeScript and React; and building backend services in Python and Go. You will work closely with product managers, designers, researchers, and the rendering team.
Key responsibilities include:
- Design and improve the agent's tools, prompts, and context to enhance planning, error recovery, and multi-turn conversation handling
- Build evaluation frameworks and offline replays of real user sessions to measure quality
- Run experiments with feature flags and analyze results; review user sessions and agent traces to identify failure points
- Develop user-facing features in TypeScript/React and backend services/workflows in Python/Go
- Partner cross-functionally with product, design, research, and rendering teams
HeyGen trains its own models in-house, and you will work directly with those models and the researchers who build them—an experience few applied AI roles offer. This is not a role limited to calling third-party APIs; you will influence model behavior and architecture.
REQUIREMENTS:
- Shipped an AI agent in production where a model calls tools, plans across multiple steps, and acts on results (chat wrappers or prompt demos alone are insufficient)
- Experience building evals or test sets for model behavior; ability to distinguish real improvements from noise and healthy skepticism of results
- 2+ years of professional software engineering experience
- Fluent in TypeScript and React (frontend) and Python (backend)
- Strong product sense: you use what you build, care about user experience, and analyze how customers actually use the product
- Ownership mindset: you identify the most important problems, fix them, and communicate changes without waiting for detailed specs
- Computer science degree or equivalent professional experience
NICE TO HAVE:
- Experience fine-tuning models
- Background with code-to-video models (HTML, animation scripts)
- Prior work on video agent models
