Applied AI Engineer, Video Agent
About this role
The problem this hire will solve is the challenge of turning complex inputs like prompts or documents into finished videos at scale. The Video Agent team is focused on building the core technology that automates this process, allowing users to generate various types of video content efficiently. This involves creating a multi-step AI agent capable of planning storyboards, scripting, selecting assets, and executing necessary tool calls to produce a final video. The role exists to enhance this agent's capability to precisely fulfill user requests and to rigorously measure that improvement through data.
This position is central to the end-to-end ownership of the Video Agent system, spanning from initial concept through to deployment and iteration. Engineers in this role will be responsible for improving the agent’s planning logic, building evaluation frameworks, and shipping user-facing enhancements. The ideal candidate thrives in an environment where they can directly influence the product, decide the boundary between model-driven and deterministic code, and take full ownership of features from inception to launch.
Success in this role requires a blend of deep technical execution across the stack and strong product intuition. It suits an engineer who enjoys deep system design, continuous experimentation, and translating user feedback into tangible improvements for an AI-powered product.
What you will do
- The role involves designing and building the core Video Agent, including defining the necessary tools, prompts, and context for the agent to operate effectively.
- Engineers will focus on improving the agent's planning capabilities, error recovery mechanisms, and its ability to handle complex, multi-turn conversations.
- A key responsibility is measuring the quality of the agent's output by building evaluation frameworks and running offline replays of real user sessions.
- The role requires shipping new features by building user-facing components using TypeScript and React.
- Engineers will develop backend services and long-running workflows utilizing Python and Go.
- This position involves making critical decisions on which aspects of the functionality should be handled by the AI model versus deterministic code.
- The individual will partner with product managers, designers, and research teams to align technical solutions with product goals.
What you need
- Candidates must have experience building and shipping an AI agent that involves a model calling tools and planning across several steps.
- Experience with a chat wrapper or prompt demonstration alone is insufficient; production experience detailing production issues and fixes is required.
- A strong understanding of how to quantify AI system improvement, including building evaluation sets and distinguishing real gains from noise.
- The candidate must be a strong full-stack engineer proficient in TypeScript and React for the frontend and Python for the backend.
- Demonstrated product sense is necessary, as the role requires understanding how users interact with the built system and caring about the user experience.
- The ideal candidate takes ownership, proactively identifying problems, fixing them, and communicating changes without waiting for detailed specifications.
- A minimum of two years of professional experience is required for this role.
Nice to have
- Fine-tuned models
- Worked on code-to-video models
- Worked on video agent models in general
- Show an agent you’ve built
Benefits
- Equity in a high-growth AI company
- 401K Match
- Health
- Dental
- Vision
- Flexible PTO
Skills and tech stack
Questions candidates ask
What is the remote policy?
The job is listed with locations in San Francisco, Palo Alto, Los Angeles, and Toronto, suggesting a presence in these areas. Specific details regarding remote work are not provided in the posting.
What is the compensation range?
The salary for this role ranges from $180,000 to $240,000 annually.
What is the primary tech stack used?
The frontend development utilizes TypeScript and React, while the backend services and long-running workflows are built using Python and Go.
What is the team structure like?
The team owns the agent end-to-end, covering the prompts, tools, backend services, and the product screens where users interact with the agent.