We're looking for a founding research engineer to build the creative intelligence layer behind Ponder with us.
Ponder is building the world's most intelligent agent for creative work.
This entails:
- Designing and shipping agentic systems that reason over large amounts of videos (audio, visuals, timeline state)
- Building self-improving agents that learn from usage, feedback, and structured evaluation
- Developing retrieval, memory, and multimodal perception systems
- Creating benchmarking pipelines that measure narrative quality, coherence, and creative judgment
- Improving tool reliability, long-horizon planning, and latency in production
You've:
- Shipped complex agentic systems in production
- Built with multimodal models (text, vision, video, audio)
- Worked deeply with embeddings, retrieval systems, and structured planning
- Designed evaluation loops that drive measurable improvement over time
- Ideally trained or fine-tuned models (SFT, preference tuning, RL, or similar methods)
You should be great at:
- Turning ambiguous creative problems into measurable systems
- Balancing research depth with product velocity