Founding AI Engineer - Hera
Role details
Job location
Tech stack
Job description
This is a founding AI engineering role working directly with the founders to build and optimize LLM-powered code generation systems for professional animation. The team is small and elite - every hire is expected to contribute at a very high level and push the company forward quickly. Future growth could include Staff AI Engineer or Head of AI as the team expands., TypeScript, React.js, Next.js, PostgreSQL, GCP/AWS, and a proprietary motion graphics engine. The team prioritizes learning velocity and end-to-end delivery.
Requirements
Do you have experience in TypeScript?, Do you have a Master's degree?, * 3+ years of experience in AI engineering, LLM systems, or related engineering roles
- Strong experience building LLM-powered systems - particularly code generation or programmatic output from natural language
- Experience designing and implementing agentic pipelines - multi-step reasoning, tool calling, autonomous task execution
- Ability to build production-ready ML/AI systems and pipelines
- Experience in fast-paced startup environments or shipping production software
- Strong programming and problem-solving skills
- Based in Berlin or able to work onsite in New York or San Francisco
Benefits & conditions
- Build and optimize LLM-powered code generation systems that produce professional animations following motion design principles - easing, non-linear movement, and visual hierarchy
- Design and implement agentic pipelines that chain multi-step reasoning and code generation to translate natural language prompts into editable animation code
- Develop evaluation frameworks to measure animation code quality using both automated metrics and expert motion design feedback
- Collaborate with motion designers and marketers to translate professional design knowledge into training data, prompts, and evaluation criteria
- Design and implement fine-tuning pipelines to teach models domain-specific animation styles and brand guidelines
- Create feedback loops between user preferences, expert input, and model performance to continuously improve generation quality
- Research and prototype new approaches for controllable, style-consistent code generation for motion graphics